conda create --name cads3 python=3.10.16

!python --version

Python 3.10.16

# Import the package `subprocess` to run shell commands and capture its output
import subprocess

# Run the pip list command to get the installed packages in the format `package==version` with --format=freeze
result = subprocess.run(["pip", "list", "--format=freeze"], 
                        capture_output=True, 
                        text=True)      # Return output as a string (instead of bytes)

# Split the output by lines and store it in a list
package_list = result.stdout.splitlines()

# Calculate the maximum length of the package-version strings for alignment of the output
max_package_length = max(len(pkg) for pkg in package_list)

# Print the packages in a nicely left-aligned format (5 per line)
print("List of installed packages:\n")

for i in range(0, len(package_list), 5):
    line = ""
    for pkg in package_list[i:i + 5]:
        # Align package-version pairs in columns
        line += f"{pkg.ljust(max_package_length)}" + "\t"
    print(line)

List of installed packages:

absl-py==2.1.0                      	asttokens==3.0.0                    	astunparse==1.6.3                   	catboost==1.2.7                     	certifi==2024.12.14                 	
charset-normalizer==3.4.0           	colorama==0.4.6                     	comm==0.2.2                         	contourpy==1.3.1                    	cycler==0.12.1                      	
debugpy==1.8.11                     	decorator==5.1.1                    	exceptiongroup==1.2.2               	executing==2.1.0                    	flatbuffers==24.3.25                	
fonttools==4.55.3                   	gast==0.6.0                         	google-pasta==0.2.0                 	graphviz==0.20.3                    	grpcio==1.68.1                      	
h5py==3.12.1                        	idna==3.10                          	importlib_metadata==8.5.0           	ipykernel==6.29.5                   	ipython==8.31.0                     	
jedi==0.19.2                        	Jinja2==3.1.5                       	joblib==1.4.2                       	jupyter_client==8.6.3               	jupyter_core==5.7.2                 	
keras==3.7.0                        	kiwisolver==1.4.7                   	libclang==18.1.1                    	lightgbm==4.5.0                     	Markdown==3.7                       	
markdown-it-py==3.0.0               	MarkupSafe==3.0.2                   	matplotlib==3.10.0                  	matplotlib-inline==0.1.7            	mdurl==0.1.2                        	
ml-dtypes==0.4.1                    	namex==0.0.8                        	nest-asyncio==1.6.0                 	numpy==1.26.4                       	opt_einsum==3.4.0                   	
optree==0.13.1                      	packaging==24.2                     	pandas==2.2.3                       	parso==0.8.4                        	patsy==1.0.1                        	
pexpect==4.9.0                      	pickleshare==0.7.5                  	pillow==11.0.0                      	pip==24.3.1                         	platformdirs==4.3.6                 	
plotly==5.24.1                      	prompt_toolkit==3.0.48              	protobuf==5.29.2                    	psutil==6.1.0                       	ptyprocess==0.7.0                   	
pure_eval==0.2.3                    	Pygments==2.18.0                    	pyparsing==3.2.0                    	python-dateutil==2.9.0.post0        	pytz==2024.2                        	
pywin32==310                        	pyzmq==26.2.0                       	requests==2.32.3                    	rich==13.9.4                        	scikit-learn==1.5.2                 	
scipy==1.14.1                       	seaborn==0.13.2                     	setuptools==75.6.0                  	six==1.17.0                         	stack-data==0.6.3                   	
statsmodels==0.14.4                 	tenacity==9.0.0                     	tensorboard==2.18.0                 	tensorboard-data-server==0.7.2      	tensorflow==2.18.0                  	
tensorflow_intel==2.18.0            	tensorflow-io-gcs-filesystem==0.31.0	termcolor==2.5.0                    	threadpoolctl==3.5.0                	tornado==6.4.2                      	
traitlets==5.14.3                   	typing_extensions==4.12.2           	tzdata==2024.2                      	urllib3==2.3.0                      	wcwidth==0.2.13                     	
Werkzeug==3.1.3                     	wheel==0.45.1                       	wrapt==1.17.0                       	xgboost==2.1.3                      	zipp==3.21.0

import os
# Set environment variable to ignore all warnings
os.environ['PYTHONWARNINGS'] = 'ignore'

# Standard Python libraries for data analysis, scientific computing, and plotting
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import time
import re
import random
from scipy.stats import uniform, loguniform

# Setting display options for pandas and matploblib
pd.set_option('display.max_columns', None)
COLOR_LIGHT, COLOR_DARK = '#849CBE', '#00548A'

# Scikit-learn modules for preprocessing and machine learning models
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split, RandomizedSearchCV
from sklearn.metrics import roc_auc_score

# Gradient boosting frameworks and logistic regression
from catboost import CatBoostClassifier
from xgboost import XGBClassifier
from lightgbm import LGBMClassifier
import statsmodels.formula.api as smf

# Tensorflow and Keras for artificial neural networks
import tensorflow as tf
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Dense, Input, Embedding, Flatten, Concatenate, Dropout
from tensorflow.keras.utils import set_random_seed

# Ensure reproducibility
RANDOM_SEED = 42
set_random_seed(RANDOM_SEED)

# # Prevent TensorFlow From Fully Allocating GPU Memory
# # Ref: https://www.tensorflow.org/guide/gpu#limiting_gpu_memory_growth
# gpus = tf.config.list_physical_devices('GPU')
# if gpus:
#     try:
#     # Currently, memory growth needs to be the same across GPU
#         for gpu in gpus:
#             tf.config.experimental.set_memory_growth(gpu, True)
#         logical_gpus = tf.config.list_logical_devices('GPU')
#         print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")
#     except RuntimeError as e:
#         # Memory growth must be set before GPUs have been initialized
#         print(e)

# Load the dataset into a pandas DataFrame without interpreting any strings as NaN
df_raw = pd.read_csv("../02_tidy_data/diabetic_data_bin.csv", # UPDATE this path if needed
                     na_filter=False) # Ensures all values are read as-is (no auto-NaN)

# Display the dimensions of the dataset (rows, columns)
print(f"Input Dataset dimensions (rows, columns): {df_raw.shape}")

# Display a random sample of 7 rows from the dataset to inspect data
df_raw.sample(n=7, random_state=RANDOM_SEED)

Input Dataset dimensions (rows, columns): (47751, 51)

# Find columns whose names end with "_id"
id_columns = [col for col in df_raw.columns if col.endswith("id")]

# Convert their types to 'object'
df_raw[id_columns] = df_raw[id_columns].astype('object')

# Format column names and dtypes
dtypes_list = [f"{col:<25}: {str(dtype):<10}" for col, dtype in df_raw.dtypes.items()]

# Print the column types in a nicely left-aligned format (5 per line)
print("List of data features and their types:\n")

for i in range(0, len(dtypes_list), 5):
    print(" ".join(dtypes_list[i:i+5]))

List of data features and their types:

race                     : object     gender                   : object     age                      : object     weight                   : object     admission_type_id        : object    
discharge_disposition_id : object     admission_source_id      : object     time_in_hospital         : int64      payer_code               : object     medical_specialty        : object    
num_lab_procedures       : int64      num_procedures           : int64      num_medications          : int64      number_outpatient        : int64      number_emergency         : int64     
number_inpatient         : int64      diag_1                   : object     diag_2                   : object     diag_3                   : object     number_diagnoses         : int64     
max_glu_serum            : object     A1Cresult                : object     metformin                : object     repaglinide              : object     nateglinide              : object    
chlorpropamide           : object     glimepiride              : object     acetohexamide            : object     glipizide                : object     glyburide                : object    
tolbutamide              : object     pioglitazone             : object     rosiglitazone            : object     acarbose                 : object     miglitol                 : object    
troglitazone             : object     tolazamide               : object     examide                  : object     citoglipton              : object     insulin                  : object    
glyburide-metformin      : object     glipizide-metformin      : object     glimepiride-pioglitazone : object     metformin-rosiglitazone  : object     metformin-pioglitazone   : object    
change                   : object     diabetesMed              : object     TARGET                   : int64      group_diag_1             : object     group_diag_2             : object    
group_diag_3             : object

# Count the occurrences of each unique value in the 'TARGET' column
target_counts = df_raw['TARGET'].value_counts()

# Calculate the total number of entries
total_entries = len(df_raw['TARGET'])

# Define a custom formatting function for the pie chart labels
def custom_autopct(pct):
    absolute = int(pct/100.*total_entries)
    return "{:.2f}%\n({:d} entries)".format(pct, absolute)

# Create a pie chart to visualize the distribution of classes in the 'TARGET' column
pie_chart, _, autotexts = plt.pie(target_counts, labels=target_counts.index.map({0: 'Class 0', 1: 'Class 1'}),
                                  autopct=custom_autopct, startangle=90,
                                  colors=[COLOR_LIGHT, COLOR_DARK], wedgeprops={'edgecolor': 'none'})

# Set the color of the autopct texts to white
for autotext in autotexts:
    autotext.set_color('white')

# Plot the pie chart
plt.title('Distribution of the TARGET Variable')
plt.show()

# Identify categorical features by data type 'object'
categorical_features = list(df_raw.select_dtypes(include=['object']).columns)

# Compute the number of unique values for each categorical feature and reset the index to make 'Feature Names' a column
categorical_feature_counts = df_raw[categorical_features].nunique().reset_index()

# Set the column names to 'Categorical Feature' and 'Number of Unique Values'
categorical_feature_counts.columns = ['Categorical Feature', 'Number of Unique Values']

# Sort the DataFrame by 'Number of Unique Values' in descending order
categorical_feature_counts.sort_values(by='Number of Unique Values', ascending=False, inplace=True)

# Display the resulting DataFrame
display(categorical_feature_counts.style.hide(axis='index'))

# Define a function that returns a DataFrame with features containing missing values
# and the percentage of missing values for each of these features
def missing_values_percentage(data):
    # Calculate percentages and create a DataFrame
    nan_percentages = data.isna().sum() * 100 / len(data)
    missing_values_df = pd.DataFrame({
        'Categorical Feature': nan_percentages.index, 
        'Missing Percent': nan_percentages.values
    })
    # Format the 'Missing Percent' column as a percentage string with two decimal places
    missing_values_df['Missing Percent'] = missing_values_df['Missing Percent'].apply(lambda x: f'{x:.2f}%')
    # Filter out features without missing values and sort by percentage
    missing_values_df = missing_values_df[missing_values_df['Missing Percent'] != '0.00%']
    missing_values_df.sort_values(by='Missing Percent', ascending=False, inplace=True)
    return missing_values_df

# Calculate the proportion of missing values for the categorical features
missing_values_df = missing_values_percentage(df_raw[categorical_features])

# Display the resulting DataFrame without the index using the new recommended Styler.hide method
missing_values_df_styled = missing_values_df.style.hide(axis='index')
display(missing_values_df_styled)

# # Replace missing values with a new category '__missingValue__' in all categorical columns
# df_raw[categorical_features] = df_raw[categorical_features].fillna('__missingValue__')

# Constants for data split ratios
TRAIN_RATIO = 0.7
VAL_TEST_RATIO = 0.5  # Splitting the remaining 30% equally into validation and test

# Separate the features (X) and the target (y)
X_raw = df_raw.drop(columns=['TARGET'], axis=1)
y = df_raw['TARGET']

# Split the dataset into a training set and a combined validation and test set
X_raw_train, X_raw_val_test, y_train, y_val_test = train_test_split(X_raw, y, train_size=TRAIN_RATIO, random_state=RANDOM_SEED)

# Further split the combined validation and test set into separate validation and test sets
X_raw_val, X_raw_test, y_val, y_test = train_test_split(X_raw_val_test, y_val_test, test_size=VAL_TEST_RATIO, random_state=RANDOM_SEED)

# Verify the dimensions of the training, validation, and test sets by displaying (rows, columns)
print(f"Training set dimensions (rows, columns):   {X_raw_train.shape}")
print(f"Validation set dimensions (rows, columns):  {X_raw_val.shape}")
print(f"Test set dimensions (rows, columns):        {X_raw_test.shape}")

Training set dimensions (rows, columns):   (33425, 50)
Validation set dimensions (rows, columns):  (7163, 50)
Test set dimensions (rows, columns):        (7163, 50)

# Calculate TARGET mean and std for each set
stats = pd.DataFrame({
    'Mean': [y_train.mean(), y_val.mean(), y_test.mean()],
    'Standard Deviation': [y_train.std(), y_val.std(), y_test.std()]
}, index=['Train', 'Validation', 'Test']).round(4)

print(stats)

              Mean  Standard Deviation
Train       0.1314              0.3378
Validation  0.1312              0.3377
Test        0.1319              0.3384

# Start timer to calculate the running time of training the CatBoost model
start_time = time.time()

# Fit the CatBoostClassifier on the training data
CB1 = CatBoostClassifier(eval_metric='AUC', random_seed=RANDOM_SEED)
CB1.fit(X_raw_train, y_train, cat_features=categorical_features, logging_level='Silent')

# Calculate and print the running time in seconds
elapsed_time_CB1 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_CB1:.2f}")

Elapsed time (sec): 157.66

# Display all model hyperparameters in a DataFrame
hyperparams_list = [(k, v) for k, v in CB1.get_all_params().items()]
hyperparams_df = pd.DataFrame(hyperparams_list, columns=['Hyperparameter', 'Value'])
display(hyperparams_df.style.hide(axis='index'))

# Extract the 15 most important features and create a Series for plotting
feature_importances = pd.Series(CB1.feature_importances_, index=X_raw_train.columns).nlargest(15)

# Plot the feature importances
feature_importances.plot(kind='barh', color=COLOR_DARK)
plt.gca().invert_yaxis()
plt.title('Top 15 Feature Importances in CatBoost Model')
plt.xlabel('Importance')
plt.ylabel('Feature')
plt.show()

# Calculate the AUC score using the model's prediction probabilities for the positive class
auc_CB1 = roc_auc_score(y_val, CB1.predict_proba(X_raw_val)[:, 1])

# Print the validation AUC
print(f'The validation AUC of the CatBoost model with standard hyperparameters is: {auc_CB1:.6f}')

The validation AUC of the CatBoost model with standard hyperparameters is: 0.707713

# # Identify numerical features
# numerical_features = list(X_raw_train.select_dtypes(exclude=['object']).columns)

# Copy the Input Data
df_pre = df_raw.copy()

# # Replace missing values for numeric features by median value from the training data
# for col in numerical_features:
#     median_value = X_raw_train[col].median()
#     df_pre[col] = df_pre[col].fillna(median_value)

# Check for remaining missing values:
num_missing = df_pre.isnull().sum().sum()
print(f"\nNumber of missing values after imputation: {num_missing}")

Number of missing values after imputation: 0

# # Replace very rare values
# print("\nGender_Frequency: ", df_pre.CODE_GENDER.value_counts())
# df_pre.loc[df_pre["CODE_GENDER"] != "M", "CODE_GENDER"] = "F"

# Constants for data split ratios
TRAIN_RATIO = 0.7
VAL_TEST_RATIO = 0.5  # Splitting the remaining 30% equally into validation and test

# Separate the features (X) and the target (y) after preprocessing
X_pre = df_pre.drop(['TARGET'], axis=1)
y_pre = df_pre['TARGET']

# Split the preprocessed dataset into a training set and a combined validation and test set
X_pre_train, X_pre_val_test, y_pre_train, y_pre_val_test = train_test_split(X_pre, y_pre, train_size=TRAIN_RATIO, random_state=RANDOM_SEED)

# Further split the combined validation and test set into separate validation and test sets
X_pre_val, X_pre_test, y_pre_val, y_pre_test = train_test_split(X_pre_val_test, y_pre_val_test, test_size=VAL_TEST_RATIO, random_state=RANDOM_SEED)

# Verify the dimensions of the training, validation, and test sets by displaying (rows, columns)
print(f"Preprocessed training set dimensions (rows, columns):   {X_pre_train.shape}")
print(f"Preprocessed validation set dimensions (rows, columns): {X_pre_val.shape}")
print(f"Preprocessed test set dimensions (rows, columns):       {X_pre_test.shape}")

Preprocessed training set dimensions (rows, columns):   (33425, 50)
Preprocessed validation set dimensions (rows, columns): (7163, 50)
Preprocessed test set dimensions (rows, columns):       (7163, 50)

# Combine features and target variable into a single dataframe for formula-based modeling
Xy_pre_train = X_pre_train.copy()
Xy_pre_train['TARGET'] = y_pre_train 

# Start timer to calculate the running time of training the logistic regression model
start_time = time.time()

# Define the logistic regression model using statsmodels' formula API and fit it to the training data
LR1 = smf.logit(formula="TARGET ~ discharge_disposition_id + group_diag_1 + number_inpatient \
                         + admission_source_id + num_lab_procedures + age + \
                                 num_medications + number_diagnoses", 
                data=Xy_pre_train).fit()

# Calculate and print the running time in seconds
elapsed_time_LR1 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_LR1:.2f}")

# Display a summary of the logistic regression model results
print(LR1.summary())

Optimization terminated successfully.
         Current function value: 0.365714
         Iterations 8
Elapsed time (sec): 1.27
                           Logit Regression Results                           
==============================================================================
Dep. Variable:                 TARGET   No. Observations:                33425
Model:                          Logit   Df Residuals:                    33378
Method:                           MLE   Df Model:                           46
Date:                Tue, 13 May 2025   Pseudo R-squ.:                 0.05995
Time:                        10:43:32   Log-Likelihood:                -12224.
converged:                       True   LL-Null:                       -13004.
Covariance Type:            nonrobust   LLR p-value:                1.047e-296
===================================================================================================
                                      coef    std err          z      P>|z|      [0.025      0.975]
---------------------------------------------------------------------------------------------------
Intercept                          -4.0687      0.721     -5.644      0.000      -5.482      -2.656
discharge_disposition_id[T.2]       0.5727      0.096      5.947      0.000       0.384       0.761
discharge_disposition_id[T.3]       0.6122      0.052     11.748      0.000       0.510       0.714
discharge_disposition_id[T.4]       0.5608      0.171      3.273      0.001       0.225       0.897
discharge_disposition_id[T.5]       1.2109      0.112     10.854      0.000       0.992       1.430
discharge_disposition_id[T.6]       0.2691      0.055      4.869      0.000       0.161       0.377
discharge_disposition_id[T.7]       0.0973      0.225      0.432      0.666      -0.344       0.539
discharge_disposition_id[T.8]       0.4815      0.488      0.987      0.324      -0.475       1.437
discharge_disposition_id[T.15]      2.8252      0.501      5.643      0.000       1.844       3.806
discharge_disposition_id[T.18]      0.2945      0.088      3.352      0.001       0.122       0.467
discharge_disposition_id[T.22]      1.4496      0.084     17.294      0.000       1.285       1.614
discharge_disposition_id[T.23]     -1.1954      0.460     -2.598      0.009      -2.097      -0.294
discharge_disposition_id[T.24]     -0.5908      1.040     -0.568      0.570      -2.629       1.447
discharge_disposition_id[T.25]      0.0319      0.182      0.175      0.861      -0.324       0.388
discharge_disposition_id[T.28]      1.9884      0.289      6.892      0.000       1.423       2.554
group_diag_1[T.Diabetes]            0.0427      0.068      0.625      0.532      -0.091       0.177
group_diag_1[T.Digestive]          -0.1658      0.066     -2.503      0.012      -0.296      -0.036
group_diag_1[T.Genitourinary]      -0.1996      0.082     -2.435      0.015      -0.360      -0.039
group_diag_1[T.Injury]             -0.2360      0.068     -3.470      0.001      -0.369      -0.103
group_diag_1[T.Musculoskeletal]    -0.3221      0.080     -4.036      0.000      -0.478      -0.166
group_diag_1[T.Neoplasms]          -0.1262      0.082     -1.539      0.124      -0.287       0.034
group_diag_1[T.Other]              -0.1623      0.052     -3.122      0.002      -0.264      -0.060
group_diag_1[T.Respiratory]        -0.3503      0.060     -5.814      0.000      -0.468      -0.232
group_diag_1[T.n.a.]                0.6518      0.909      0.717      0.473      -1.130       2.433
admission_source_id[T.2]           -0.2494      0.159     -1.569      0.117      -0.561       0.062
admission_source_id[T.3]            0.5929      0.303      1.959      0.050      -0.000       1.186
admission_source_id[T.4]           -0.3855      0.096     -4.036      0.000      -0.573      -0.198
admission_source_id[T.5]           -0.5668      0.204     -2.777      0.005      -0.967      -0.167
admission_source_id[T.6]           -0.2210      0.117     -1.891      0.059      -0.450       0.008
admission_source_id[T.7]            0.0393      0.042      0.944      0.345      -0.042       0.121
admission_source_id[T.8]           -0.1709      1.097     -0.156      0.876      -2.321       1.979
admission_source_id[T.9]           -0.2962      0.475     -0.623      0.533      -1.228       0.636
admission_source_id[T.17]           0.0189      0.073      0.260      0.795      -0.124       0.162
admission_source_id[T.20]           0.7139      0.306      2.334      0.020       0.114       1.313
age[T.[10-20)]                      1.0586      0.754      1.403      0.160      -0.420       2.537
age[T.[20-30)]                      1.3588      0.732      1.857      0.063      -0.075       2.793
age[T.[30-40)]                      1.2057      0.725      1.663      0.096      -0.215       2.626
age[T.[40-50)]                      1.2979      0.721      1.799      0.072      -0.116       2.712
age[T.[50-60)]                      1.2224      0.721      1.696      0.090      -0.190       2.635
age[T.[60-70)]                      1.3950      0.721      1.936      0.053      -0.018       2.808
age[T.[70-80)]                      1.5193      0.721      2.108      0.035       0.107       2.932
age[T.[80-90)]                      1.4326      0.721      1.986      0.047       0.019       2.846
age[T.[90-100)]                     1.1406      0.728      1.567      0.117      -0.286       2.567
number_inpatient                    0.5878      0.025     23.090      0.000       0.538       0.638
num_lab_procedures                  0.0027      0.001      2.902      0.004       0.001       0.005
num_medications                     0.0036      0.002      1.654      0.098      -0.001       0.008
number_diagnoses                    0.0557      0.010      5.755      0.000       0.037       0.075
===================================================================================================

# Evaluate the performance of the logistic regression model on the validation data
# and calculate the Area Under the Curve (AUC) for the ROC
auc_LR1 = roc_auc_score(y_pre_val, LR1.predict(X_pre_val))

# Print the validation AUC
print(f'The validation AUC of the logistic regression model is: {auc_LR1:.6f}')

The validation AUC of the logistic regression model is: 0.680391

# # Start timer to calculate the running time of training the CatBoost model with median replacement
# start_time = time.time()

# # Initialize a CatBoostClassifier and fit it to the preprocessed training data
# CB2 = CatBoostClassifier(eval_metric='AUC', random_seed=RANDOM_SEED)
# CB2.fit(X_pre_train, y_train, cat_features=categorical_features, logging_level='Silent')

# # Calculate and print the running time in seconds
# elapsed_time_CB2 = time.time() - start_time
# print(f"Elapsed time (sec): {elapsed_time_CB2:.2f}")

# # Evaluate the performance of the CatBoost model with median replacement on the
# # validation data and calculate the Area Under the Curve (AUC) for the ROC
# auc_CB2 = roc_auc_score(y_val, CB2.predict_proba(X_pre_val)[:,1])
# print(f'The validation AUC of the CatBoost model with median replacement is:   {auc_CB2:.6f}')
# print(f'The validation AUC of the CatBoost model with minimum replacement was: {auc_CB1:.6f}')

# List of selected categorical features to analyze
selected_categorical_features = ['group_diag_1', 'group_diag_2', 'group_diag_3', 
                                 'age', 'medical_specialty', 'payer_code', 'discharge_disposition_id']

# Number of rows/columns for the subplot grid
n_cols = 2  # doubled to fit count and percentage plots side by side
n_rows = len(selected_categorical_features)  # one row for each feature

# Set up the matplotlib figure
plt.figure(figsize=(n_cols * 9, n_rows * 5))

# Loop through the number of categorical features
for idx, feature in enumerate(selected_categorical_features):
    # Create a crosstab for stacked bar plot structure
    ctab = pd.crosstab(df_pre[feature], df_pre['TARGET'])
    
    # (1st column) add a new subplot iteratively for count values
    ax1 = plt.subplot(n_rows, n_cols, idx * n_cols + 1)
    
    # Create a stacked bar plot for count values
    ctab.plot(kind="bar", stacked=True, color=[COLOR_LIGHT, COLOR_DARK], edgecolor="none", ax=ax1)
    
    # Additional plot settings for count subplot
    ax1.set_title(f'Stacked bar plot of {feature}')
    ax1.set_xlabel(feature)
    ax1.set_ylabel('Count')
    ax1.set_yscale('log') # Set y-axis to log base 10
    ax1.legend(title='TARGET', loc='center left', bbox_to_anchor=(1, 0.5))
    plt.xticks(rotation=45, ha='right')
    
    # (2nd column) add a new subplot iteratively for percentage values
    ax2 = plt.subplot(n_rows, n_cols, idx * n_cols + 2)
        
    # Normalize the crosstab by row and multiply by 100 to convert to percentages
    ctab_normalized = ctab.div(ctab.sum(axis=1), axis=0) * 100
    
    # Create a stacked bar plot for percentage values
    ctab_normalized.plot(kind="bar", stacked=True, color=[COLOR_LIGHT, COLOR_DARK], edgecolor="none", ax=ax2)
    
    # Additional plot settings for percentage subplot
    ax2.set_title(f'Stacked bar plot of {feature} (in %)')
    ax2.set_xlabel(feature)
    ax2.set_ylabel('Percentage [%]')
    ax2.legend().remove()
    plt.xticks(rotation=45, ha='right')
    
    # Make yticks be in percentages for the percentage subplot
    ax2.yaxis.set_major_formatter(plt.FuncFormatter(lambda x, loc: "{:.0f}%".format(int(x))))
    ax2.set_ylim(0, 100)
    
# Preventing subplots from being too close to each other and display the final plot
plt.tight_layout()
plt.show()

# List of selected numerical features to analyze
selected_numerical_features = ['number_outpatient', 'number_inpatient', 'num_procedures', 'num_lab_procedures', 
                               'num_medications', 'time_in_hospital', 'number_emergency', 'number_diagnoses']

# number of plots, set up the subplot grid
n_plots = len(selected_numerical_features)
n_cols = 2  # adjust the number of columns based on your preference
n_rows = (n_plots + n_cols - 1) // n_cols

# set up the matplotlib figure
plt.figure(figsize=(n_cols * 5, n_rows * 4))

# loop through the list of numerical features
for idx, feature in enumerate(selected_numerical_features):
    ax = plt.subplot(n_rows, n_cols, idx + 1)
    
    # Create a color mapping based on TARGET values assuming values are 0 and 1
    color_mapping = {0: COLOR_LIGHT, 1: COLOR_DARK}
    
    class_vals = df_pre['TARGET'].unique()
    for val in class_vals:
        subset = df_pre[df_pre['TARGET'] == val]
        
        # Draw the KDE for each class
        sns.kdeplot(subset[feature], ax=ax, label=f'TARGET {val}', color=color_mapping[val], linewidth=2.5)
    
    # additional plot settings
    ax.set_title(f'Distribution of {feature}')
    ax.set_xlabel(f'{feature}')
    ax.set_ylabel('Density')
    ax.legend(title='TARGET')

# Preventing subplots from being too close to each other and display the final plot
plt.tight_layout()
plt.show()

# Output the dimensions of the dataset prior to dummy encoding
print("Shape of the dataset before dummy encoding: ", df_pre.shape) 

# Perform dummy encoding on categorical features, removing the first level to avoid dummy variable trap
df = pd.get_dummies(df_pre, drop_first=True)

# Output the dimensions of the dataset after dummy encoding to show the change
print("Shape of the dataset after dummy encoding: ", df.shape)

Shape of the dataset before dummy encoding:  (47751, 51)
Shape of the dataset after dummy encoding:  (47751, 2249)

# Change columns names (LightGBM does not support special JSON characters in feature names)
new_names = {col: re.sub(r'[^A-Za-z0-9_]+', '', col) for col in df.columns}
new_n_list = list(new_names.values())
new_names = {col: f'{new_col}_{i}' if new_col in new_n_list[:i] else new_col for i, (col, new_col) in enumerate(new_names.items())}
df = df.rename(columns=new_names)
df.head(1)

# Generate the samples once again

# Split dataset in 70% training, 15% validation and 15% test and devide into feature matrix x and label y
X = df.drop(['TARGET'], axis=1)
y = df.TARGET
X_train, X_valtest, y_train, y_valtest = train_test_split(X, y, train_size=0.70, random_state=RANDOM_SEED)
X_val, X_test, y_val, y_test = train_test_split(X_valtest, y_valtest, test_size=0.50, random_state=RANDOM_SEED)
print('Sample sizes for training, validation and test: ', X_train.shape, X_val.shape, X_test.shape)

Sample sizes for training, validation and test:  (33425, 2248) (7163, 2248) (7163, 2248)

# Generate list: feature names  
feature_names=list(X.columns)

# Feature scaling using StandardScaler based on training data distributions 
scaler = StandardScaler()
scaler.fit(X_train)
X_train = pd.DataFrame(scaler.transform(X_train), columns = feature_names)
X_val = pd.DataFrame(scaler.transform(X_val), columns = feature_names)

# Repeat CB with one-hot-encoded and scaled data
start_time = time.time()

CB4 = CatBoostClassifier(eval_metric='AUC', random_seed=RANDOM_SEED) 
CB4.fit(X_train, y_train, logging_level='Silent')

# Calculate and print the running time in seconds
elapsed_time_CB4 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_CB4:.2f}")

Elapsed time (sec): 22.69

# CatBoost: Plot most the top 20 most important features
pd.Series(CB4.feature_importances_, index=X_train.columns).nlargest(20).plot(kind='barh', color=COLOR_DARK).invert_yaxis()
plt.show()

# Calculate and dispay the model AUC  
auc_CB4 = roc_auc_score(y_val, CB4.predict_proba(X_val)[:,1])
print('The validation AUC of the CatBoost model after encoding and scaling is:   {:.6f}'.format(auc_CB4))
print('The validation AUC of the CatBoost model before encoding and scaling was: {:.6f}'.format(auc_CB1))

The validation AUC of the CatBoost model after encoding and scaling is:   0.702311
The validation AUC of the CatBoost model before encoding and scaling was: 0.707713

print(f"Elapsed time of the CatBoost model after encoding and scaling (sec):  {elapsed_time_CB4:.2f}")
print(f"Elapsed time of the CatBoost model before encoding and scaling (sec): {elapsed_time_CB1:.2f}")

Elapsed time of the CatBoost model after encoding and scaling (sec):  22.69
Elapsed time of the CatBoost model before encoding and scaling (sec): 157.66

non_event_factor = 2 # thus the event rate should be 33.3% (instead of 13.1%) 

# Save complete training data
X_train_all = X_train.copy(deep=True)
y_train_all = y_train.copy(deep=True)

# Subsampling: enhance event rate to 1:nov_event_factor
Xs = pd.concat([X_train.reset_index(drop=True), y_train.reset_index(drop=True)], axis=1)
Xs = pd.concat([Xs[Xs['TARGET']==1], Xs[Xs['TARGET']==0].sample(
           n = non_event_factor * len(Xs[Xs['TARGET']==1] ), random_state=RANDOM_SEED)],axis=0).sort_index(ascending=True)

# Over-write complete training data with subsample
X_train = Xs.drop(columns=["TARGET"])
y_train = Xs["TARGET"]

print("\nSize of training data set after subsampling: ", X_train.shape)
print("\nTARGET-distribution after subsampling:\n", y_train.value_counts())
print("\nReduction of data volume (%): ", round((len(y_train)/len(y_train_all)-1)*100))

Size of training data set after subsampling:  (13176, 2248)

TARGET-distribution after subsampling:
 TARGET
0    8784
1    4392
Name: count, dtype: int64

Reduction of data volume (%):  -61

# Repeat CB with subsampled data
start_time = time.time()

CB5 = CatBoostClassifier(eval_metric='AUC', random_seed=RANDOM_SEED) 
CB5.fit(X_train, y_train, logging_level='Silent')

# Calculate and print the running time in seconds
elapsed_time_CB5 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_CB5:.2f}")

Elapsed time (sec): 21.06

# Calculate and dispay the model AUC 
auc_CB5 = roc_auc_score(y_val, CB5.predict_proba(X_val)[:,1])
print('The validation AUC of the CatBoost model with subsampled data is: {:.6f}'.format(auc_CB5))

The validation AUC of the CatBoost model with subsampled data is: 0.699061

# Calculate relative performance
print("Reduction of training time (%): ", round((elapsed_time_CB5/elapsed_time_CB4-1)*100))
print("Improvment in AUC score (%): ", round((auc_CB5/auc_CB4-1)*100, 2))

Reduction of training time (%):  -7
Improvment in AUC score (%):  -0.46

# Data structures for model names, AUC, as well as training time
mname = []
mauc = []
mtrainingtime = []

dict = {'Model name': mname,'AUC': mauc, 'Training time': mtrainingtime}

# Store name and AUC of prevoisly fitted models 
mname.append("CB1_quick")
mauc.append(auc_CB1)
mtrainingtime.append(elapsed_time_CB1)
mname.append("LR1_select")
mauc.append(auc_LR1)
mtrainingtime.append(elapsed_time_LR1)
mname.append("CB4_OHE")
mauc.append(auc_CB4)
mtrainingtime.append(elapsed_time_CB4)
mname.append("CB5_after subsampling")
mauc.append(auc_CB5)
mtrainingtime.append(elapsed_time_CB5)

def plot_model_performance(d, x1, x2, t): 
    df_eval = pd.DataFrame(d)
    
    # Create a figure with 1 row and 2 columns
    plt.figure(figsize=(20, 6))

    # First subplot (left): Model AUCs
    ax1 = plt.subplot(1, 2, 1)
    ax1.set_title(" Model evaluation (AUC): " + t)
    sns.barplot(data = df_eval, x = "AUC", y = "Model name", color = COLOR_DARK)
    ax1.set_xlim(x1, x2)
    
    # Second subplot (right): Model training times
    ax2 = plt.subplot(1, 2, 2)
    ax2.set_title(" Model training time in second: " + t)
    sns.barplot(data = df_eval, x = "Training time", y = "Model name", color = COLOR_DARK)
    ax2.set_ylabel("")
    
    plt.tight_layout()
    plt.show()

plot_model_performance(dict, 0.66, 0.72, "Validation")

# Define the hyperparameters grid to search
param_grid_CB ={'learning_rate': loguniform(0.01, 0.05), 
                'depth':         [4,5,6,7]}

# Define variables for results reports
sel_params_CB = ['param_learning_rate','param_depth','mean_test_score','rank_test_score']

# CatBoostClassifier: Hyper parameter optimization with RandomizedSearchCV
tic = time.time()

CB_rs = RandomizedSearchCV(CatBoostClassifier(iterations=1000, eval_metric='AUC'), param_grid_CB,
                           cv=4, n_iter=10, n_jobs=-1, random_state=RANDOM_SEED, scoring='roc_auc')
CB_rs.fit(X_train, y_train, logging_level='Silent')

# Print the runtime as well as the five best hyper parameter constellations
print("time (sec):" + "%6.0f" % (time.time() - tic))
print("Best hyper parameters:",CB_rs.best_params_)
pd.DataFrame(CB_rs.cv_results_)[sel_params_CB].sort_values("rank_test_score").head()

time (sec):   240
Best hyper parameters: {'depth': 6, 'learning_rate': 0.020941182421473574}

# CatBoostClassifier: Fit new model on all folds (with best parameters) 
start_time = time.time()
CB6 = CatBoostClassifier(**CB_rs.best_params_, iterations=1000, eval_metric='AUC')
CB6.fit(X_train, y_train, logging_level='Silent')

# Calculate and print the running time in seconds
elapsed_time_CB6 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_CB6:.2f}")

Elapsed time (sec): 18.48

# Calculate and dispay the model AUC 
auc_CB6 = roc_auc_score(y_val, CB6.predict_proba(X_val)[:,1])
print('The validation AUC of the CatBoost model with tuned hyperparameters is: {:.6f}'.format(auc_CB6))
print('Comparison: The val. AUC of the CatBoost model with subsampled data is: {:.6f}'.format(auc_CB5))

The validation AUC of the CatBoost model with tuned hyperparameters is: 0.700497
Comparison: The val. AUC of the CatBoost model with subsampled data is: 0.699061

# Parameter search range for LightGBM
param_grid_LGB = {'learning_rate':    loguniform(0.001, 0.02), 
                  'num_leaves':       [25,32,40,50],
                  'subsample':        uniform(0,1),
                  'colsample_bytree': uniform(0,1),
                  'verbose':          [-1]}

# Create a list of variables for displaying the cross-validation result
sel_params_LGB = ['param_learning_rate', 'param_num_leaves', 'param_subsample',
                  'param_colsample_bytree', 'mean_test_score', 'rank_test_score']

# LGBMClassifier: Hyperparameter optimization with RandomizedSearchCV
tic = time.time()

LGB_rs = RandomizedSearchCV(LGBMClassifier(n_estimators=1000),
                            param_grid_LGB, cv=4, n_iter=20, scoring='roc_auc',
                            n_jobs=-1, random_state=RANDOM_SEED)
LGB_rs.fit(X_train, y_train)

# Output the runtime and best parameters (as well as the runner-ups)
print("Elapsed time (sec):" + "%6.0f" % (time.time() - tic))
print("Best Parameters:", LGB_rs.best_params_)
pd.DataFrame(LGB_rs.cv_results_)[sel_params_LGB].sort_values("rank_test_score").head()

Elapsed time (sec):   360
Best Parameters: {'colsample_bytree': 0.24102546602601171, 'learning_rate': 0.007743661011598839, 'num_leaves': 50, 'subsample': 0.4951769101112702, 'verbose': -1}

# LGBMClassifier: Create a new model on all folds using the best parameters
start_time = time.time()
LGB = LGBMClassifier(**LGB_rs.best_params_, n_estimators=1000)    
LGB.fit(X_train, y_train)

# Calculate and print the running time in seconds
elapsed_time_LGB = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_LGB:.2f}")

Elapsed time (sec): 6.51

# Calculate and dispay the model AUC 
auc_LGB = roc_auc_score(y_val, LGB.predict_proba(X_val)[:,1])
print('The validation AUC of the LightGBM model with tuned hyperparameters is: {:.6f}'.format(auc_LGB))

The validation AUC of the LightGBM model with tuned hyperparameters is: 0.705074

# Parameter search range for XGBoost
param_grid_XGB = {'learning_rate':    loguniform(0.005, 0.03), 
                  'max_depth':        [6,7,8,9],
                  'subsample':        uniform(0,1),
                  'colsample_bytree': uniform(0,1)}

# Create a list of variables for displaying the cross-validation result
sel_params_XGB = ['param_learning_rate', 'param_max_depth', 'param_subsample',
                  'param_colsample_bytree', 'mean_test_score', 'rank_test_score']

# XGBClassifier: Hyperparameter optimization with RandomizedSearchCV
tic = time.time()

XGB_rs = RandomizedSearchCV(XGBClassifier(n_estimators=1000, tree_method="hist"),
                            param_grid_XGB, cv=4, n_iter=10, scoring="roc_auc",
                            n_jobs=-1, random_state=RANDOM_SEED)
XGB_rs.fit(X_train, y_train)

# Output the runtime and best parameters (as well as the runner-ups)
print("Elapsed time (sec):" + "%6.0f" % (time.time() - tic))
print("Best Parameters:", XGB_rs.best_params_)
pd.DataFrame(XGB_rs.cv_results_)[sel_params_XGB].sort_values("rank_test_score").head()

Elapsed time (sec):   587
Best Parameters: {'colsample_bytree': 0.18182496720710062, 'learning_rate': 0.006945227129221507, 'max_depth': 9, 'subsample': 0.6116531604882809}

# XGBClassifier: Create a new model on all folds using the best parameters
start_time = time.time()
XGB = XGBClassifier(**XGB_rs.best_params_, n_estimators=1000, tree_method="hist", eval_metric="auc")    
XGB.fit(X_train, y_train)

# Calculate and print the running time in seconds
elapsed_time_XGB = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_XGB:.2f}")

Elapsed time (sec): 68.60

# Calculate and display the model AUC 
auc_XGB = roc_auc_score(y_val, XGB.predict_proba(X_val)[:,1])
print('The validation AUC of the XGBoost model with tuned hyperparameters is: {:.6f}'.format(auc_XGB))

The validation AUC of the XGBoost model with tuned hyperparameters is: 0.703716

# Load the dataset `icd9_data.csv` into a pandas DataFrame
icd9 = pd.read_csv("../02_tidy_data/icd9_data.csv", # UPDATE this path if needed
                   na_filter=False) # no `NA` values in our prepared dataset

# Display the first three rows from the dataset to inspect data
icd9.head(n=3)

# Extract the joining variables of the original dataset
diag_cols = ['diag_1', 'diag_2', 'diag_3']

# Columns to retain from the icd9 data
cols_to_retain = ['group', 'L2code', 'category']

# Define a helper function to merge each diagnosis `diag_col` with the respective retained column `icd9_col`
def merge_diag_column(df, icd9_col, diag_col):
    return df.merge(icd9[['code', icd9_col]], 
                    left_on=diag_col, right_on='code', how='left')\
             .rename(columns={
                 icd9_col: f'{icd9_col}_{diag_col}'
             })\
             .drop(columns='code')
             
# Copy the original dataset and drop the column `group` for consistency
df_pre2 = df_raw.copy().drop(columns=[f'group_{diag}' for diag in diag_cols])

# Merge the datasets
for icd9_col in cols_to_retain:
    for diag_col in diag_cols:
        df_pre2 = merge_diag_column(df_pre2, icd9_col, diag_col)

# Replace missing values with the same logic as the dataset `icd9`
for i in range(1, 4):
    df_pre2[f'group_diag_{i}'] = df_pre2[f'group_diag_{i}'].fillna('n.a.')
    df_pre2[f'L2code_diag_{i}'] = df_pre2[f'L2code_diag_{i}'].fillna('n.a.')
    df_pre2[f'category_diag_{i}'] = df_pre2[f'category_diag_{i}'].fillna('Not Available')
    
# Display the first three rows of the merged dataset
df_pre2.head(n=3)

# Filter for readmitted patients
readmitted_df = df_pre2[df_pre2['TARGET'] == 1]

# Stack category_diag_1, category_diag_2, category_diag_3 into a single column
category_cols = [f'category_{diag}' for diag in diag_cols]
readmitted_stacked = readmitted_df[category_cols].melt(value_name='category')

# Count the frequency of each category across all diagnosis levels
target_category_counts = readmitted_stacked['category'].value_counts()

# Sort in descending order
target_category_counts = target_category_counts.sort_values(ascending=False)

# Display the top 3 most frequent categories
print("The top three categories across all three diagnosis levels:")
display(target_category_counts.reset_index().head(3).style.hide(axis='index'))

The top three categories across all three diagnosis levels:

# Filter for readmitted patients
readmitted_df = df_pre2[df_pre2['TARGET'] == 1]

# Initialize dictionary to hold counts
target_category_counts = {}

# Loop over each diagnosis column
for diag in diag_cols:
    # Count value frequencies for this column
    counts = readmitted_df[f'category_{diag}'].value_counts()
    
    # Sort in descending order and store in dictionary
    target_category_counts[diag] = counts.sort_values(ascending=False)
    
    # Display the top three most frequent categories
    print(f"The top three categories of {diag}:")
    display(target_category_counts[diag].reset_index()
            .rename(columns={f'category_{diag}': 'category'}).head(3).style.hide(axis='index'))

The top three categories of diag_1:

The top three categories of diag_2:

The top three categories of diag_3:

# Add the required new column `sum_target_cg` for each diagnosis
for col in diag_cols:
    # A dictionary storing the absolute proportion of each category in relation to readmissions
    sum_target_cg_dict = target_category_counts[diag].to_dict()

    df_pre2[f'sum_target_cg_{col}'] = df_pre2[f'category_{col}'].map(sum_target_cg_dict)
    
    # Fill it with the value zero if the category is not found among readmitted patients
    df_pre2[f'sum_target_cg_{col}'] = df_pre2[f'sum_target_cg_{col}'].fillna(0).astype('int64')

threshold = 20

# Function to add a new column adjusting the value of each `category`
def create_hdiag(diag, category, group):
    # Set the count again to zero if the category is not found among readmitted patients
    count = target_category_counts[diag].get(category, 0)
    
    # Set the value of `hdiag` to category if the category appears more than 20 times 
    # and the string desired otherwise
    value = category
    if not count > threshold:
        value = f"o.{group}"
    
    # Trim to max 30 characters
    return str(value)[:30]

# Create a new column `hdiag_x` for each diagnosis `diag_x`
for diag in diag_cols:
    df_pre2[f'h{diag}'] = df_pre2.apply(
        lambda row: create_hdiag(diag, row[f'category_{diag}'], row[f'group_{diag}']), 
        axis=1)

# Function to verify the content of the column `hdiag` for a few random rows
def test_hdiag_logic(num_checks=3):
    passed = 0
    failed = 0
    failed_rows = []

    sample_rows = df_pre2.sample(n=num_checks, random_state=RANDOM_SEED)
    
    for idx, row in sample_rows.iterrows():
        for diag in diag_cols:
            category = row[f'category_{diag}']
            group = row[f'group_{diag}']
            hdiag = row[f'h{diag}']
            
            # Get value counts of the corresponding category
            count = target_category_counts[diag].get(category, 0)
            
            # expected value of `hdiag`
            expected = category if count > threshold else f"o.{group}"
            expected = str(expected)[:30]
            
            try:
                assert hdiag == expected
                passed += 1
            except AssertionError:
                failed += 1
                failed_rows.append({
                    "index": idx,
                    "diag": diag,
                    "category": category,
                    "group": group,
                    "count": count,
                    "hdiag": hdiag,
                    "expected": expected
                })

    print(f"✅ {passed} checks passed")
    if failed:
        print(f"❌ {failed} checks failed. Here are the details:")
        for row in failed_rows:
            print(row)

# Verify the column `diag` by checking two sample rows
test_hdiag_logic(num_checks=2)

✅ 6 checks passed

# Define the columns to be adjusted and their replacement values
categorical_features_to_adjust = {
    'discharge_disposition_id': 99,
    'medical_specialty': 'ZZ',
    'payer_code': 'ZZ',
    'admission_type_id': 99, # Assuming 'admission_type' is 'admission_type_id'
    'age': '[0-20)'
}

# Apply replacements based on frequency
for col, replacement in categorical_features_to_adjust.items():
    # Count frequency of each value
    value_counts = df_pre2[col].value_counts()
    
    # Identify values below threshold
    rare_values = value_counts[value_counts < threshold].index
    
    # Replace rare values
    df_pre2[col] = df_pre2[col].apply(lambda x: replacement if x in rare_values else x)

# Display updated values and frequencies
for col in categorical_features_to_adjust.keys():
    print(f"\nUpdated value counts for '{col}':")
    display(df_pre2[col].value_counts().reset_index().style.hide(axis='index'))

Updated value counts for 'discharge_disposition_id':

Updated value counts for 'medical_specialty':

Updated value counts for 'payer_code':

Updated value counts for 'admission_type_id':

Updated value counts for 'age':

# Function to display relative frequencies for categorical features
def plot_relative_frequencies(features, fig_width, fig_height):
    for feature in features:
        plt.figure(figsize=(fig_width, fig_height))
        feature_counts = df_pre2[feature].value_counts(normalize=True) * 100
        
        sns.barplot(
            x=feature_counts.index.astype(str), y=feature_counts.values,
            color=COLOR_DARK
        )
        
        plt.title(f"Distribution of {feature} as Percentage of Total")
        plt.ylabel("Percentage (%)")
        plt.yscale('log') # Set y-axis to log base 10
        plt.xlabel(feature)
        plt.xticks(rotation=45, ha='right')
        
        plt.tight_layout()
        plt.show()

# Plot each categorical feature as a bar chart showing relative frequencies
plot_relative_frequencies(categorical_features_to_adjust.keys(), 10, 5)
plot_relative_frequencies([f'h{diag}' for diag in diag_cols], 18, 6)

# List of categorical columns to encode
categorical_features = list(categorical_features_to_adjust.keys()) + [f'h{diag}' for diag in diag_cols]

# Store the dataset before encoding as a backup
df_before_embeddings = df_pre2.copy()

# Perform label encoding manually (alternatively with LabelEncoder from sklearn)
for col in categorical_features:
    unique_vals = df_pre2[col].astype(str).unique()
    val_to_int = {val: idx for idx, val in enumerate(sorted(unique_vals))} # `sorted(unique_vals)` ensures consistent ordering.
    df_pre2[col] = df_pre2[col].astype(str).map(val_to_int)

# Display three random rows
df_pre2.sample(n=3, random_state=RANDOM_SEED)

# Numerical columns to scale
numerical_features = ['number_inpatient', 'num_lab_procedures', 'number_diagnoses', 'num_medications']

# Compute means and stds before scaling (for comparison)
before_scaling = df_pre2[numerical_features].describe()

# Manual standardization
for col in numerical_features:
    mean = df_pre2[col].mean()
    std = df_pre2[col].std()
    df_pre2[col] = (df_pre2[col] - mean) / std

# Compute means and stds after scaling
after_scaling = df_pre2[numerical_features].describe()

# Show comparison side-by-side
print("Before Scaling:\n", before_scaling)
print("\nAfter Scaling:\n", after_scaling)

Before Scaling:
        number_inpatient  num_lab_procedures  number_diagnoses  num_medications
count      47751.000000        47751.000000      47751.000000     47751.000000
mean           0.138259           42.356746          7.102092        15.559716
std            0.532646           19.809284          2.046605         8.516024
min            0.000000            1.000000          1.000000         1.000000
25%            0.000000           30.000000          5.000000        10.000000
50%            0.000000           44.000000          8.000000        14.000000
75%            0.000000           56.000000          9.000000        20.000000
max           12.000000          132.000000         16.000000        81.000000

After Scaling:
        number_inpatient  num_lab_procedures  number_diagnoses  num_medications
count      4.775100e+04        4.775100e+04      4.775100e+04     4.775100e+04
mean      -2.142744e-17        4.999735e-17     -1.761811e-16    -4.419409e-17
std        1.000000e+00        1.000000e+00      1.000000e+00     1.000000e+00
min       -2.595701e-01       -2.087746e+00     -2.981568e+00    -1.709685e+00
25%       -2.595701e-01       -6.237856e-01     -1.027112e+00    -6.528535e-01
50%       -2.595701e-01        8.295371e-02      4.387304e-01    -1.831507e-01
75%       -2.595701e-01        6.887303e-01      9.273444e-01     5.214034e-01
max        2.226948e+01        4.525315e+00      4.347643e+00     7.684370e+00

# Plot the distribution of numerical features before and after scaling
for col in numerical_features:
    fig, axes = plt.subplots(1, 2, figsize=(12, 4))
    sns.histplot(df_before_embeddings[col], ax=axes[0], kde=True, color=COLOR_LIGHT)
    axes[0].set_title(f"{col} - Before Scaling")
    
    sns.histplot(df_pre2[col], ax=axes[1], kde=True, color=COLOR_DARK)
    axes[1].set_title(f"{col} - After Scaling")
    
    plt.tight_layout()
    plt.show()

# Create the dataset consisting of the features processed above together with the TARGET
df2 = df_pre2[categorical_features + numerical_features + ['TARGET']]

# Split the dataset into training, validation and test data
# Constants for data split ratios
TRAIN_RATIO = 0.7
VAL_TEST_RATIO = 0.5  # Splitting the remaining 30% equally into validation and test

# Separate the features (X) and the target (y)
X2 = df2.drop(columns=['TARGET'], axis=1)
y2 = df2['TARGET']

# Split the dataset into a training set and a combined validation and test set
X2_train, X2_val_test, y2_train, y2_val_test = train_test_split(X2, y2, train_size=TRAIN_RATIO, random_state=RANDOM_SEED)

# Further split the combined validation and test set into separate validation and test sets
X2_val, X2_test, y2_val, y2_test = train_test_split(X2_val_test, y2_val_test, test_size=VAL_TEST_RATIO, random_state=RANDOM_SEED)

# Verify the dimensions of the training, validation, and test sets by displaying (rows, columns)
print(f"Training set dimensions (rows, columns):   {X2_train.shape}")
print(f"Validation set dimensions (rows, columns):  {X2_val.shape}")
print(f"Test set dimensions (rows, columns):        {X2_test.shape}")

Training set dimensions (rows, columns):   (33425, 12)
Validation set dimensions (rows, columns):  (7163, 12)
Test set dimensions (rows, columns):        (7163, 12)

def build_model(layers, dropout_rate, optimizer):
    
    # Define input layers and embedding layers
    inputs = {}
    embeddings = {}
    embedding_dim = 2  # 2D embeddings

    for col in categorical_features:
        # Number of unique feature values
        num_unique = df2[col].nunique()
        # Input layer
        inputs[col] = Input(shape=(1,), name=f"{col}_input", dtype='int64')
        
        # Embedding layer
        embeddings[col] = Embedding(input_dim=num_unique, output_dim=embedding_dim, name=f"{col}_embedding")(inputs[col])
        embeddings[col] = Flatten(name=f"{col}_flatten")(embeddings[col])

    # Add input for numerical features
    inputs['numerical'] = Input(shape=(len(numerical_features),), name="numerical_input")

    # Concatenate all features: categorical embeddings + numerical input
    all_features = list(embeddings.values()) + [inputs['numerical']]
    x = Concatenate(name="concat_layer")(all_features)
    
    if isinstance(layers, int): # 1 hidden layer
        x = Dense(layers, activation='relu', name=f'dense')(x)
        x = Dropout(dropout_rate, name=f'dropout')(x)
    else:
        for i, nodes in enumerate(layers):
            x = Dense(nodes, activation='relu', name=f'dense_{i+1}')(x)
            x = Dropout(dropout_rate, name=f'dropout_{i+1}')(x)

    # Output layer
    output = Dense(1, activation='sigmoid', name='output')(x)

    # Define and compile our neural network model
    ANN = Model(inputs=list(inputs.values()), outputs=output)
    ANN.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['auc'])

    return ANN

# Categorical features
X2_train_cat = [X2_train[col].values for col in categorical_features]
X2_val_cat = [X2_val[col].values for col in categorical_features]
X2_test_cat = [X2_test[col].values for col in categorical_features]

# Numerical features (assumed already scaled)
X2_train_num = X2_train[numerical_features].values
X2_val_num = X2_val[numerical_features].values
X2_test_num = X2_test[numerical_features].values

# Combine and split
X2_train_ANN = X2_train_cat + [X2_train_num]
X2_val_ANN = X2_val_cat + [X2_val_num]
X2_test_ANN = X2_test_cat + [X2_test_num]

# Hyperparameter search range 
param_grid = {
    'layers': [(10), (20), (30), (10, 5), (20, 5), (30, 5), (20, 10), (30, 10), (30,20,5), (30,20,10)],
    'dropout_rate': [0.1, 0.3, 0.5],
    'optimizer':  ['adam'], 
    'batch_size': [32, 64],
    'epochs': [10, 20]
}

# Define EarlyStopping callback
early_stopping = tf.keras.callbacks.EarlyStopping(
    monitor='val_auc', # Watch validation auc
    patience=5, # Number of epochs with no improvement after which training will be stopped
    restore_best_weights=True # Restore model weights from the epoch with the best value of the monitored quantity
)

# Artificial neural network: Hyperparameter optimization with RandomizedSearchCV
tic = time.time()

results = []

# Ensure reproducibility
set_random_seed(RANDOM_SEED)

# Start the hyperparameter tuning
for i in range(20):  # Try 20 random configurations
    params = {
        k: random.choice(v)
        for k, v in param_grid.items()
    }

    # print(f"Trial {i+1}: {params}")
    ANN = build_model(params['layers'], params['dropout_rate'], params['optimizer'])
    
    # Fit the neural network defined in PT5-2 a)
    history = ANN.fit(
        X2_train_ANN, y2_train,
        validation_data=(X2_val_ANN, y2_val),
        epochs=params['epochs'],
        callbacks=[early_stopping],
        batch_size=params['batch_size'],
        verbose=0
    )

    val_loss = min(history.history["val_loss"])
    val_auc = max(history.history["val_auc"])

    results.append((params, val_loss, val_auc))

# Sort by best validation auc
results.sort(key=lambda x: x[2], reverse=True)

# Output the runtime and best parameters (as well as the runner-ups)
print("Elapsed time (sec):" + "%6.0f" % (time.time() - tic))
print("Best: %f using %s" % (results[0][2], results[0][0]))
result_df = pd.DataFrame(results, columns=['params', 'val_loss', 'val_auc'])
pd.concat([result_df['params'].apply(pd.Series), result_df[['val_loss', 'val_auc']]], axis=1).head(12)

Elapsed time (sec):   372
Best: 0.692090 using {'layers': (30, 10), 'dropout_rate': 0.3, 'optimizer': 'adam', 'batch_size': 32, 'epochs': 20}

# Artificial neural network: Create a new model on all folds using the best parameters
start_time = time.time()

# Reset the seed after neural network tuning to ensure reproducibility
set_random_seed(RANDOM_SEED)

# Build the neural network model with the best hyperparameters
ANN = build_model((10), 0.1, 'adam')

# Fit the model
ANN_history = ANN.fit(
    X2_train_ANN, y2_train,
    validation_data=(X2_val_ANN, y_val),
    epochs=10,
    callbacks=[early_stopping],
    batch_size=32,
    verbose=1
)

# Calculate and print the running time in seconds
elapsed_time_ANN = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_ANN:.2f}")

Epoch 1/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 6s 3ms/step - auc: 0.5288 - loss: 0.5071 - val_auc: 0.6794 - val_loss: 0.3631
Epoch 2/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 2ms/step - auc: 0.6379 - loss: 0.3789 - val_auc: 0.6843 - val_loss: 0.3608
Epoch 3/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - auc: 0.6598 - loss: 0.3717 - val_auc: 0.6831 - val_loss: 0.3598
Epoch 4/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 2ms/step - auc: 0.6680 - loss: 0.3687 - val_auc: 0.6803 - val_loss: 0.3601
Epoch 5/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - auc: 0.6775 - loss: 0.3656 - val_auc: 0.6801 - val_loss: 0.3600
Epoch 6/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 2ms/step - auc: 0.6795 - loss: 0.3651 - val_auc: 0.6798 - val_loss: 0.3603
Epoch 7/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - auc: 0.6832 - loss: 0.3637 - val_auc: 0.6782 - val_loss: 0.3602
Elapsed time (sec): 20.94

# Plot training and validation loss
plt.plot(ANN_history.history['loss'], label='Training Loss', color=COLOR_LIGHT)
plt.plot(ANN_history.history['val_loss'], label='Validation Loss', color=COLOR_DARK)
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training vs Validation Loss')
plt.legend()
plt.grid(True)

plt.tight_layout()
plt.show()

# Calculate and display the model validation loss
val_loss = ANN.evaluate(X2_val_ANN, y2_val, verbose=0)[0]
print(f"Final Validation Loss: {val_loss:.4f}")

Final Validation Loss: 0.3608

# Calculate and display the model validation AUC 
auc_ANN = ANN.evaluate(X2_val_ANN, y2_val, verbose=0)[1]
print('The validation AUC of the neural network model with tuned hyperparameters is: {:.6f}'.format(auc_ANN))

The validation AUC of the neural network model with tuned hyperparameters is: 0.684344

# Clone the netork from PT5-2 a)
ANN_clone = tf.keras.models.clone_model(ANN)
ANN_clone.compile(optimizer='adam', loss='binary_crossentropy', metrics=['auc'])  # recompile

# Fit the network once more
ANN_history_clone = ANN_clone.fit(
    X2_train_ANN, y2_train,
    validation_data=(X2_val_ANN, y2_val),
    epochs=10,
    callbacks=[early_stopping],
    batch_size=32,
    verbose=1
)

# Calculate and output the validation loss
val_loss_clone = ANN.evaluate(X2_val_ANN, y2_val, verbose=0)[0]
print(f"Final Validation Loss: {val_loss_clone:.4f}")

Epoch 1/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 5s 3ms/step - auc: 0.5361 - loss: 0.4957 - val_auc: 0.6773 - val_loss: 0.3632
Epoch 2/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - auc: 0.6450 - loss: 0.3755 - val_auc: 0.6817 - val_loss: 0.3614
Epoch 3/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - auc: 0.6644 - loss: 0.3700 - val_auc: 0.6817 - val_loss: 0.3605
Epoch 4/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - auc: 0.6747 - loss: 0.3667 - val_auc: 0.6797 - val_loss: 0.3602
Epoch 5/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - auc: 0.6821 - loss: 0.3644 - val_auc: 0.6785 - val_loss: 0.3604
Epoch 6/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - auc: 0.6868 - loss: 0.3628 - val_auc: 0.6771 - val_loss: 0.3608
Epoch 7/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - auc: 0.6847 - loss: 0.3630 - val_auc: 0.6759 - val_loss: 0.3611
Epoch 8/10
1045/1045 ━━━━━━━━━━━━━━━━━━━━ 3s 2ms/step - auc: 0.6896 - loss: 0.3616 - val_auc: 0.6754 - val_loss: 0.3609
Final Validation Loss: 0.3608

# Function to extract embedding weights
def get_embedding_weights(model, feature):
    layer = model.get_layer(f"{feature}_embedding")
    return layer.get_weights()[0]

# Prepare plot grid
n = len(categorical_features)
fig, axes = plt.subplots(n, 2, figsize=(12, n * 4))
fig.suptitle("Embedding Comparison per Feature", fontsize=16, y=1.02)

# Loop through features and plot
for i, feature in enumerate(categorical_features):
    emb1 = get_embedding_weights(ANN, feature)
    emb2 = get_embedding_weights(ANN_clone, feature)

    # Original model
    axes[i, 0].scatter(emb1[:, 0], emb1[:, 1], c=COLOR_LIGHT)
    axes[i, 0].set_title(f"{feature} - Original")
    axes[i, 0].grid(True)
    
    # Cloned model
    axes[i, 1].scatter(emb2[:, 0], emb2[:, 1], c=COLOR_DARK)
    axes[i, 1].set_title(f"{feature} - Cloned")
    axes[i, 1].grid(True)
    
plt.tight_layout()
plt.show()

# Loop through features and extend dataset with its embedding vectors
for i, feature in enumerate(categorical_features):
    # Get embedding weights and map each category to its embedding
    emb_weights = get_embedding_weights(ANN, feature)
    embeddings = {label: emb_weights[label] for label in range(emb_weights.shape[0])}

    # Replace the categorical column with its embedding columns
    embedding_df = df2[feature].map(embeddings)
    embedding_df = pd.DataFrame(embedding_df.tolist(),
                                index=df2.index,
                                columns=[f"{feature}_emb_{i}" for i in range(emb_weights.shape[1])])
    
    # Join embeddings to the original dataset
    df2 = pd.concat([df2, embedding_df], axis=1)
    
    # Drop the original categorical columns if now embedded
    df2.drop(columns=[feature], inplace=True)

# Split the dataset into training, validation and test data
# Constants for data split ratios
TRAIN_RATIO = 0.7
VAL_TEST_RATIO = 0.5  # Splitting the remaining 30% equally into validation and test

# Separate the features (X) and the target (y)
X2 = df2.drop(columns=['TARGET'], axis=1)
y2 = df2['TARGET']

# Split the dataset into a training set and a combined validation and test set
X2_train, X2_val_test, y2_train, y2_val_test = train_test_split(X2, y2, train_size=TRAIN_RATIO, random_state=RANDOM_SEED)

# Further split the combined validation and test set into separate validation and test sets
X2_val, X2_test, y2_val, y2_test = train_test_split(X2_val_test, y2_val_test, test_size=VAL_TEST_RATIO, random_state=RANDOM_SEED)

# Verify the dimensions of the training, validation, and test sets by displaying (rows, columns)
print(f"Training set dimensions (rows, columns):   {X2_train.shape}")
print(f"Validation set dimensions (rows, columns):  {X2_val.shape}")
print(f"Test set dimensions (rows, columns):        {X2_test.shape}")

Training set dimensions (rows, columns):   (33425, 20)
Validation set dimensions (rows, columns):  (7163, 20)
Test set dimensions (rows, columns):        (7163, 20)

# Combine features and target variable into a single dataframe for formula-based modeling
Xy2_train = X2_train.copy()
Xy2_val = X2_val.copy()

Xy2_train['TARGET'] = y2_train 
Xy2_val['TARGET'] = y2_val

# Start timer to calculate the running time of training the logistic regression model
start_time = time.time()

# Define the logistic regression model using statsmodels' formula API and fit it to the training data
predictors = "+".join([col for col in X2_train.columns if not re.search(r"hdiag_2|hdiag_3", col)])
LR2 = smf.logit(formula=f"TARGET ~ {predictors}", data=Xy2_train).fit()

# Calculate and print the running time in seconds
elapsed_time_LR2 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_LR2:.2f}")

# Display a summary of the logistic regression model results
print(LR2.summary())

Optimization terminated successfully.
         Current function value: 0.362905
         Iterations 7
Elapsed time (sec): 0.12
                           Logit Regression Results                           
==============================================================================
Dep. Variable:                 TARGET   No. Observations:                33425
Model:                          Logit   Df Residuals:                    33408
Method:                           MLE   Df Model:                           16
Date:                Tue, 13 May 2025   Pseudo R-squ.:                 0.06717
Time:                        11:19:43   Log-Likelihood:                -12130.
converged:                       True   LL-Null:                       -13004.
Covariance Type:            nonrobust   LLR p-value:                     0.000
==================================================================================================
                                     coef    std err          z      P>|z|      [0.025      0.975]
--------------------------------------------------------------------------------------------------
Intercept                         -0.4283      0.114     -3.768      0.000      -0.651      -0.206
number_inpatient                   0.2863      0.014     21.147      0.000       0.260       0.313
num_lab_procedures                 0.0295      0.019      1.565      0.118      -0.007       0.067
number_diagnoses                   0.1246      0.019      6.530      0.000       0.087       0.162
num_medications                    0.0424      0.018      2.333      0.020       0.007       0.078
discharge_disposition_id_emb_0     0.8267      0.355      2.327      0.020       0.130       1.523
discharge_disposition_id_emb_1     0.6854      0.326      2.100      0.036       0.046       1.325
medical_specialty_emb_0            0.4651      0.484      0.960      0.337      -0.484       1.415
medical_specialty_emb_1            2.8357      0.495      5.725      0.000       1.865       3.806
payer_code_emb_0                  -0.1563      0.476     -0.329      0.742      -1.088       0.776
payer_code_emb_1                  -1.1815      0.412     -2.868      0.004      -1.989      -0.374
admission_type_id_emb_0           -0.3465      0.317     -1.095      0.274      -0.967       0.274
admission_type_id_emb_1           -0.5162      0.320     -1.613      0.107      -1.143       0.111
age_emb_0                          0.3768      0.295      1.278      0.201      -0.201       0.955
age_emb_1                          0.5991      0.307      1.949      0.051      -0.003       1.201
hdiag_1_emb_0                     -1.3968      0.438     -3.188      0.001      -2.256      -0.538
hdiag_1_emb_1                     -1.0984      0.346     -3.171      0.002      -1.777      -0.420
==================================================================================================

# Evaluate the performance of the logistic regression model on the validation data
# and calculate the Area Under the Curve (AUC) for the ROC
auc_LR2 = roc_auc_score(y2_val, LR2.predict(X2_val))

# Print the validation AUC
print(f'The validation AUC of the logistic regression model is: {auc_LR2:.6f}')

The validation AUC of the logistic regression model is: 0.677810

# Start timer to calculate the running time of training the CatBoost model
start_time = time.time()

# Fit the CatBoostClassifier on the training data
CB7 = CatBoostClassifier(eval_metric='AUC', random_seed=RANDOM_SEED)
CB7.fit(X2_train, y2_train, logging_level='Silent')

# Calculate and print the running time in seconds
elapsed_time_CB7 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_CB7:.2f}")

Elapsed time (sec): 15.83

# Calculate the AUC score using the model's prediction probabilities for the positive class
auc_CB7 = roc_auc_score(y2_val, CB7.predict_proba(X2_val)[:, 1])

# Print the validation AUC
print(f'The validation AUC of the CatBoost model with standard hyperparameters is: {auc_CB7:.6f}')

The validation AUC of the CatBoost model with standard hyperparameters is: 0.673485

# Loop through features and extend dataset with its embedding vectors
for i, feature in enumerate(categorical_features):
    # Get embedding weights and map each category to its embedding
    emb_weights = get_embedding_weights(ANN, feature)
    embeddings = {label: emb_weights[label] for label in range(emb_weights.shape[0])}

    # Replace the categorical column with its embedding columns
    embedding_df = df_pre2[feature].map(embeddings)
    embedding_df = pd.DataFrame(embedding_df.tolist(),
                                index=df_pre2.index,
                                columns=[f"{feature}_emb_{i}" for i in range(emb_weights.shape[1])])
    
    # Join embeddings to the original dataset
    df_pre2 = pd.concat([df_pre2, embedding_df], axis=1)
    
    # Drop the original categorical columns if now embedded
    df_pre2.drop(columns=[feature], inplace=True)

# Split the dataset into training, validation and test data
# Constants for data split ratios
TRAIN_RATIO = 0.7
VAL_TEST_RATIO = 0.5  # Splitting the remaining 30% equally into validation and test

# Separate the features (X) and the target (y)
X2_all = df_pre2.drop(columns=['TARGET'], axis=1)
y2 = df_pre2['TARGET']

# Split the dataset into a training set and a combined validation and test set
X2_all_train, X2_all_val_test, y2_train, y2_val_test = train_test_split(X2_all, y2, train_size=TRAIN_RATIO, random_state=RANDOM_SEED)

# Further split the combined validation and test set into separate validation and test sets
X2_all_val, X2_all_test, y2_val, y2_test = train_test_split(X2_all_val_test, y2_val_test, test_size=VAL_TEST_RATIO, random_state=RANDOM_SEED)

# Verify the dimensions of the training, validation, and test sets by displaying (rows, columns)
print(f"Training set dimensions (rows, columns):   {X2_all_train.shape}")
print(f"Validation set dimensions (rows, columns):  {X2_all_val.shape}")
print(f"Test set dimensions (rows, columns):        {X2_all_test.shape}")

Training set dimensions (rows, columns):   (33425, 70)
Validation set dimensions (rows, columns):  (7163, 70)
Test set dimensions (rows, columns):        (7163, 70)

# Start timer to calculate the running time of training the CatBoost model
start_time = time.time()

# Identify categorical features by data type 'object'
cat_features = list(df_pre2.select_dtypes(include=['object']).columns)

# Fit the CatBoostClassifier on the training data
CB8 = CatBoostClassifier(eval_metric='AUC', cat_features=cat_features, random_seed=RANDOM_SEED)
CB8.fit(X2_all_train, y2_train, logging_level='Silent')

# Calculate and print the running time in seconds
elapsed_time_CB8 = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_CB8:.2f}")

Elapsed time (sec): 170.29

# Calculate the AUC score using the model's prediction probabilities for the positive class
auc_CB8 = roc_auc_score(y2_val, CB8.predict_proba(X2_all_val)[:, 1])

# Print the validation AUC
print(f'The validation AUC of the CatBoost model with standard hyperparameters is: {auc_CB8:.6f}')

The validation AUC of the CatBoost model with standard hyperparameters is: 0.701283

# Evaluate the impact of the embeddings in the models considered
# Store name and AUC of prevoisly fitted models 
mname.extend(["CB6_tuned", "LGB_tuned", "XGB_tuned", 
              "ANN_tuned", "LR2_with_embeddings", 
              "CB7_with_embeddings", "CB8_quick_with_embeddings"])
mauc.extend([auc_CB6, auc_LGB, auc_XGB, 
             auc_ANN, auc_LR2, auc_CB7, auc_CB8])
mtrainingtime.extend([elapsed_time_CB6, elapsed_time_LGB, elapsed_time_XGB, 
                      elapsed_time_ANN, elapsed_time_LR2, 
                      elapsed_time_CB7, elapsed_time_CB8])

plot_model_performance(dict, 0.66, 0.72, "Validation")

# Install AutoGluon and import required module
!pip install autogluon
from autogluon.tabular import TabularPredictor

Collecting autogluon
  Using cached autogluon-1.3.0-py3-none-any.whl.metadata (11 kB)
Collecting autogluon.core==1.3.0 (from autogluon.core[all]==1.3.0->autogluon)
  Using cached autogluon.core-1.3.0-py3-none-any.whl.metadata (12 kB)
Collecting autogluon.features==1.3.0 (from autogluon)
  Using cached autogluon.features-1.3.0-py3-none-any.whl.metadata (11 kB)
Collecting autogluon.tabular==1.3.0 (from autogluon.tabular[all]==1.3.0->autogluon)
  Using cached autogluon.tabular-1.3.0-py3-none-any.whl.metadata (14 kB)
Collecting autogluon.multimodal==1.3.0 (from autogluon)
  Using cached autogluon.multimodal-1.3.0-py3-none-any.whl.metadata (13 kB)
Collecting autogluon.timeseries==1.3.0 (from autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached autogluon.timeseries-1.3.0-py3-none-any.whl.metadata (12 kB)
Requirement already satisfied: numpy<2.3.0,>=1.25.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (1.26.4)
Requirement already satisfied: scipy<1.16,>=1.5.4 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (1.14.1)
Requirement already satisfied: scikit-learn<1.7.0,>=1.4.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (1.5.2)
Collecting networkx<4,>=3.0 (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached networkx-3.4.2-py3-none-any.whl.metadata (6.3 kB)
Requirement already satisfied: pandas<2.3.0,>=2.0.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2.2.3)
Collecting tqdm<5,>=4.38 (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached tqdm-4.67.1-py3-none-any.whl.metadata (57 kB)
Requirement already satisfied: requests in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2.32.3)
Requirement already satisfied: matplotlib<3.11,>=3.7.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (3.10.0)
Collecting boto3<2,>=1.10 (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached boto3-1.38.13-py3-none-any.whl.metadata (6.6 kB)
Collecting autogluon.common==1.3.0 (from autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached autogluon.common-1.3.0-py3-none-any.whl.metadata (11 kB)
Collecting hyperopt<0.2.8,>=0.2.7 (from autogluon.core[all]==1.3.0->autogluon)
  Using cached hyperopt-0.2.7-py2.py3-none-any.whl.metadata (1.7 kB)
Collecting ray<2.45,>=2.10.0 (from ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached ray-2.44.1-cp310-cp310-win_amd64.whl.metadata (20 kB)
Collecting pyarrow>=15.0.0 (from autogluon.core[all]==1.3.0->autogluon)
  Using cached pyarrow-20.0.0-cp310-cp310-win_amd64.whl.metadata (3.4 kB)
Requirement already satisfied: Pillow<12,>=10.0.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.multimodal==1.3.0->autogluon) (11.0.0)
Collecting torch<2.7,>=2.2 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached torch-2.6.0-cp310-cp310-win_amd64.whl.metadata (28 kB)
Collecting lightning<2.7,>=2.2 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached lightning-2.5.1.post0-py3-none-any.whl.metadata (39 kB)
Collecting transformers<4.50,>=4.38.0 (from transformers[sentencepiece]<4.50,>=4.38.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached transformers-4.49.0-py3-none-any.whl.metadata (44 kB)
Collecting accelerate<2.0,>=0.34.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached accelerate-1.6.0-py3-none-any.whl.metadata (19 kB)
Collecting jsonschema<4.24,>=4.18 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached jsonschema-4.23.0-py3-none-any.whl.metadata (7.9 kB)
Collecting seqeval<1.3.0,>=1.2.2 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached seqeval-1.2.2-py3-none-any.whl
Collecting evaluate<0.5.0,>=0.4.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached evaluate-0.4.3-py3-none-any.whl.metadata (9.2 kB)
Collecting timm<1.0.7,>=0.9.5 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached timm-1.0.3-py3-none-any.whl.metadata (43 kB)
Collecting torchvision<0.22.0,>=0.16.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached torchvision-0.21.0-cp310-cp310-win_amd64.whl.metadata (6.3 kB)
Collecting scikit-image<0.26.0,>=0.19.1 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached scikit_image-0.25.2-cp310-cp310-win_amd64.whl.metadata (14 kB)
Collecting text-unidecode<1.4,>=1.3 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached text_unidecode-1.3-py2.py3-none-any.whl.metadata (2.4 kB)
Collecting torchmetrics<1.8,>=1.2.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached torchmetrics-1.7.1-py3-none-any.whl.metadata (21 kB)
Collecting omegaconf<2.4.0,>=2.1.1 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached omegaconf-2.3.0-py3-none-any.whl.metadata (3.9 kB)
Collecting pytorch-metric-learning<2.9,>=1.3.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached pytorch_metric_learning-2.8.1-py3-none-any.whl.metadata (18 kB)
Collecting nlpaug<1.2.0,>=1.1.10 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached nlpaug-1.1.11-py3-none-any.whl.metadata (14 kB)
Collecting nltk<4.0,>=3.4.5 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached nltk-3.9.1-py3-none-any.whl.metadata (2.9 kB)
Collecting openmim<0.4.0,>=0.3.7 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached openmim-0.3.9-py2.py3-none-any.whl.metadata (16 kB)
Collecting defusedxml<0.7.2,>=0.7.1 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached defusedxml-0.7.1-py2.py3-none-any.whl.metadata (32 kB)
Requirement already satisfied: jinja2<3.2,>=3.0.3 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.multimodal==1.3.0->autogluon) (3.1.5)
Requirement already satisfied: tensorboard<3,>=2.9 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.multimodal==1.3.0->autogluon) (2.18.0)
Collecting pytesseract<0.4,>=0.3.9 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached pytesseract-0.3.13-py3-none-any.whl.metadata (11 kB)
Collecting nvidia-ml-py3<8.0,>=7.352.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached nvidia_ml_py3-7.352.0-py3-none-any.whl
Collecting pdf2image<1.19,>=1.17.0 (from autogluon.multimodal==1.3.0->autogluon)
  Using cached pdf2image-1.17.0-py3-none-any.whl.metadata (6.2 kB)
Requirement already satisfied: xgboost<3.1,>=2.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.tabular[all]==1.3.0->autogluon) (2.1.3)
Collecting fastai<2.9,>=2.3.1 (from autogluon.tabular[all]==1.3.0->autogluon)
  Using cached fastai-2.8.1-py3-none-any.whl.metadata (9.5 kB)
Requirement already satisfied: catboost<1.3,>=1.2 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.tabular[all]==1.3.0->autogluon) (1.2.7)
Collecting huggingface-hub[torch] (from autogluon.tabular[all]==1.3.0->autogluon)
  Using cached huggingface_hub-0.31.1-py3-none-any.whl.metadata (13 kB)
Collecting spacy<3.9 (from autogluon.tabular[all]==1.3.0->autogluon)
  Using cached spacy-3.8.5-cp310-cp310-win_amd64.whl.metadata (28 kB)
Collecting einops<0.9,>=0.7 (from autogluon.tabular[all]==1.3.0->autogluon)
  Using cached einops-0.8.1-py3-none-any.whl.metadata (13 kB)
Requirement already satisfied: lightgbm<4.7,>=4.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.tabular[all]==1.3.0->autogluon) (4.5.0)
Requirement already satisfied: joblib<2,>=1.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon) (1.4.2)
Collecting pytorch-lightning (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached pytorch_lightning-2.5.1.post0-py3-none-any.whl.metadata (20 kB)
Collecting gluonts<0.17,>=0.15.0 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached gluonts-0.16.1-py3-none-any.whl.metadata (9.8 kB)
Collecting statsforecast<2.0.2,>=1.7.0 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached statsforecast-2.0.1-cp310-cp310-win_amd64.whl.metadata (30 kB)
Collecting mlforecast<0.14,>0.13 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached mlforecast-0.13.6-py3-none-any.whl.metadata (12 kB)
Collecting utilsforecast<0.2.11,>=0.2.3 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached utilsforecast-0.2.10-py3-none-any.whl.metadata (7.4 kB)
Collecting coreforecast<0.0.16,>=0.0.12 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached coreforecast-0.0.15-cp310-cp310-win_amd64.whl.metadata (3.8 kB)
Collecting fugue>=0.9.0 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached fugue-0.9.1-py3-none-any.whl.metadata (18 kB)
Collecting orjson~=3.9 (from autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached orjson-3.10.18-cp310-cp310-win_amd64.whl.metadata (43 kB)
Requirement already satisfied: psutil<7.1.0,>=5.7.3 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from autogluon.common==1.3.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (6.1.0)
Requirement already satisfied: packaging>=20.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from accelerate<2.0,>=0.34.0->autogluon.multimodal==1.3.0->autogluon) (24.2)
Collecting pyyaml (from accelerate<2.0,>=0.34.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached PyYAML-6.0.2-cp310-cp310-win_amd64.whl.metadata (2.1 kB)
Collecting safetensors>=0.4.3 (from accelerate<2.0,>=0.34.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached safetensors-0.5.3-cp38-abi3-win_amd64.whl.metadata (3.9 kB)
Collecting botocore<1.39.0,>=1.38.13 (from boto3<2,>=1.10->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached botocore-1.38.13-py3-none-any.whl.metadata (5.7 kB)
Collecting jmespath<2.0.0,>=0.7.1 (from boto3<2,>=1.10->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached jmespath-1.0.1-py3-none-any.whl.metadata (7.6 kB)
Collecting s3transfer<0.13.0,>=0.12.0 (from boto3<2,>=1.10->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached s3transfer-0.12.0-py3-none-any.whl.metadata (1.7 kB)
Requirement already satisfied: graphviz in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from catboost<1.3,>=1.2->autogluon.tabular[all]==1.3.0->autogluon) (0.20.3)
Requirement already satisfied: plotly in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from catboost<1.3,>=1.2->autogluon.tabular[all]==1.3.0->autogluon) (5.24.1)
Requirement already satisfied: six in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from catboost<1.3,>=1.2->autogluon.tabular[all]==1.3.0->autogluon) (1.17.0)
Collecting datasets>=2.0.0 (from evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached datasets-3.6.0-py3-none-any.whl.metadata (19 kB)
Collecting dill (from evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached dill-0.4.0-py3-none-any.whl.metadata (10 kB)
Collecting xxhash (from evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached xxhash-3.5.0-cp310-cp310-win_amd64.whl.metadata (13 kB)
Collecting multiprocess (from evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached multiprocess-0.70.18-py310-none-any.whl.metadata (7.5 kB)
Collecting fsspec>=2021.05.0 (from fsspec[http]>=2021.05.0->evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached fsspec-2025.3.2-py3-none-any.whl.metadata (11 kB)
Requirement already satisfied: pip in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon) (24.3.1)
Collecting fastdownload<2,>=0.0.5 (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached fastdownload-0.0.7-py3-none-any.whl.metadata (5.5 kB)
Collecting fastcore<1.9,>=1.8.0 (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached fastcore-1.8.2-py3-none-any.whl.metadata (3.7 kB)
Collecting fasttransform>=0.0.2 (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached fasttransform-0.0.2-py3-none-any.whl.metadata (7.6 kB)
Collecting fastprogress>=0.2.4 (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached fastprogress-1.0.3-py3-none-any.whl.metadata (5.6 kB)
Collecting plum-dispatch (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached plum_dispatch-2.5.7-py3-none-any.whl.metadata (7.5 kB)
Collecting cloudpickle (from fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached cloudpickle-3.1.1-py3-none-any.whl.metadata (7.1 kB)
Collecting triad>=0.9.7 (from fugue>=0.9.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached triad-0.9.8-py3-none-any.whl.metadata (6.3 kB)
Collecting adagio>=0.2.4 (from fugue>=0.9.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached adagio-0.2.6-py3-none-any.whl.metadata (1.8 kB)
Collecting pydantic<3,>=1.7 (from gluonts<0.17,>=0.15.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached pydantic-2.11.4-py3-none-any.whl.metadata (66 kB)
Collecting toolz~=0.10 (from gluonts<0.17,>=0.15.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached toolz-0.12.1-py3-none-any.whl.metadata (5.1 kB)
Requirement already satisfied: typing-extensions~=4.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from gluonts<0.17,>=0.15.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon) (4.12.2)
Collecting future (from hyperopt<0.2.8,>=0.2.7->autogluon.core[all]==1.3.0->autogluon)
  Using cached future-1.0.0-py3-none-any.whl.metadata (4.0 kB)
Collecting py4j (from hyperopt<0.2.8,>=0.2.7->autogluon.core[all]==1.3.0->autogluon)
  Using cached py4j-0.10.9.9-py2.py3-none-any.whl.metadata (1.3 kB)
Requirement already satisfied: MarkupSafe>=2.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from jinja2<3.2,>=3.0.3->autogluon.multimodal==1.3.0->autogluon) (3.0.2)
Collecting attrs>=22.2.0 (from jsonschema<4.24,>=4.18->autogluon.multimodal==1.3.0->autogluon)
  Using cached attrs-25.3.0-py3-none-any.whl.metadata (10 kB)
Collecting jsonschema-specifications>=2023.03.6 (from jsonschema<4.24,>=4.18->autogluon.multimodal==1.3.0->autogluon)
  Using cached jsonschema_specifications-2025.4.1-py3-none-any.whl.metadata (2.9 kB)
Collecting referencing>=0.28.4 (from jsonschema<4.24,>=4.18->autogluon.multimodal==1.3.0->autogluon)
  Using cached referencing-0.36.2-py3-none-any.whl.metadata (2.8 kB)
Collecting rpds-py>=0.7.1 (from jsonschema<4.24,>=4.18->autogluon.multimodal==1.3.0->autogluon)
  Using cached rpds_py-0.24.0-cp310-cp310-win_amd64.whl.metadata (4.2 kB)
Collecting lightning-utilities<2.0,>=0.10.0 (from lightning<2.7,>=2.2->autogluon.multimodal==1.3.0->autogluon)
  Using cached lightning_utilities-0.14.3-py3-none-any.whl.metadata (5.6 kB)
Requirement already satisfied: contourpy>=1.0.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from matplotlib<3.11,>=3.7.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (1.3.1)
Requirement already satisfied: cycler>=0.10 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from matplotlib<3.11,>=3.7.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (0.12.1)
Requirement already satisfied: fonttools>=4.22.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from matplotlib<3.11,>=3.7.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (4.55.3)
Requirement already satisfied: kiwisolver>=1.3.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from matplotlib<3.11,>=3.7.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (1.4.7)
Requirement already satisfied: pyparsing>=2.3.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from matplotlib<3.11,>=3.7.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (3.2.0)
Requirement already satisfied: python-dateutil>=2.7 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from matplotlib<3.11,>=3.7.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2.9.0.post0)
Collecting numba (from mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached numba-0.61.2-cp310-cp310-win_amd64.whl.metadata (2.9 kB)
Collecting optuna (from mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached optuna-4.3.0-py3-none-any.whl.metadata (17 kB)
Collecting window-ops (from mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached window_ops-0.0.15-py3-none-any.whl.metadata (6.8 kB)
Collecting gdown>=4.0.0 (from nlpaug<1.2.0,>=1.1.10->autogluon.multimodal==1.3.0->autogluon)
  Using cached gdown-5.2.0-py3-none-any.whl.metadata (5.8 kB)
Collecting click (from nltk<4.0,>=3.4.5->autogluon.multimodal==1.3.0->autogluon)
  Downloading click-8.2.0-py3-none-any.whl.metadata (2.5 kB)
Collecting regex>=2021.8.3 (from nltk<4.0,>=3.4.5->autogluon.multimodal==1.3.0->autogluon)
  Using cached regex-2024.11.6-cp310-cp310-win_amd64.whl.metadata (41 kB)
Collecting antlr4-python3-runtime==4.9.* (from omegaconf<2.4.0,>=2.1.1->autogluon.multimodal==1.3.0->autogluon)
  Using cached antlr4_python3_runtime-4.9.3-py3-none-any.whl
Requirement already satisfied: colorama in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon) (0.4.6)
Collecting model-index (from openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached model_index-0.1.11-py3-none-any.whl.metadata (3.9 kB)
Collecting opendatalab (from openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached opendatalab-0.0.10-py3-none-any.whl.metadata (6.4 kB)
Requirement already satisfied: rich in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon) (13.9.4)
Collecting tabulate (from openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached tabulate-0.9.0-py3-none-any.whl.metadata (34 kB)
Requirement already satisfied: pytz>=2020.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from pandas<2.3.0,>=2.0.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2024.2)
Requirement already satisfied: tzdata>=2022.7 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from pandas<2.3.0,>=2.0.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2024.2)
Collecting filelock (from ray<2.45,>=2.10.0->ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached filelock-3.18.0-py3-none-any.whl.metadata (2.9 kB)
Collecting msgpack<2.0.0,>=1.0.0 (from ray<2.45,>=2.10.0->ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached msgpack-1.1.0-cp310-cp310-win_amd64.whl.metadata (8.6 kB)
Requirement already satisfied: protobuf!=3.19.5,>=3.15.3 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from ray<2.45,>=2.10.0->ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon) (5.29.2)
Collecting aiosignal (from ray<2.45,>=2.10.0->ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached aiosignal-1.3.2-py2.py3-none-any.whl.metadata (3.8 kB)
Collecting frozenlist (from ray<2.45,>=2.10.0->ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached frozenlist-1.6.0-cp310-cp310-win_amd64.whl.metadata (16 kB)
Collecting aiohttp>=3.7 (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached aiohttp-3.11.18-cp310-cp310-win_amd64.whl.metadata (8.0 kB)
Collecting aiohttp_cors (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached aiohttp_cors-0.8.1-py3-none-any.whl.metadata (20 kB)
Collecting colorful (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached colorful-0.5.6-py2.py3-none-any.whl.metadata (16 kB)
Collecting py-spy>=0.2.0 (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached py_spy-0.4.0-py2.py3-none-win_amd64.whl.metadata (16 kB)
Requirement already satisfied: grpcio>=1.42.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon) (1.68.1)
Collecting opencensus (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached opencensus-0.11.4-py2.py3-none-any.whl.metadata (12 kB)
Collecting prometheus_client>=0.7.1 (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached prometheus_client-0.21.1-py3-none-any.whl.metadata (1.8 kB)
Collecting smart_open (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached smart_open-7.1.0-py3-none-any.whl.metadata (24 kB)
Collecting virtualenv!=20.21.1,>=20.0.24 (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached virtualenv-20.31.2-py3-none-any.whl.metadata (4.5 kB)
Collecting tensorboardX>=1.9 (from ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB)
Requirement already satisfied: charset-normalizer<4,>=2 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from requests->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (3.4.0)
Requirement already satisfied: idna<4,>=2.5 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from requests->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (3.10)
Requirement already satisfied: urllib3<3,>=1.21.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from requests->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2.3.0)
Requirement already satisfied: certifi>=2017.4.17 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from requests->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (2024.12.14)
Collecting imageio!=2.35.0,>=2.33 (from scikit-image<0.26.0,>=0.19.1->autogluon.multimodal==1.3.0->autogluon)
  Using cached imageio-2.37.0-py3-none-any.whl.metadata (5.2 kB)
Collecting tifffile>=2022.8.12 (from scikit-image<0.26.0,>=0.19.1->autogluon.multimodal==1.3.0->autogluon)
  Using cached tifffile-2025.5.10-py3-none-any.whl.metadata (31 kB)
Collecting lazy-loader>=0.4 (from scikit-image<0.26.0,>=0.19.1->autogluon.multimodal==1.3.0->autogluon)
  Using cached lazy_loader-0.4-py3-none-any.whl.metadata (7.6 kB)
Requirement already satisfied: threadpoolctl>=3.1.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from scikit-learn<1.7.0,>=1.4.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon) (3.5.0)
Collecting spacy-legacy<3.1.0,>=3.0.11 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached spacy_legacy-3.0.12-py2.py3-none-any.whl.metadata (2.8 kB)
Collecting spacy-loggers<2.0.0,>=1.0.0 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached spacy_loggers-1.0.5-py3-none-any.whl.metadata (23 kB)
Collecting murmurhash<1.1.0,>=0.28.0 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached murmurhash-1.0.12-cp310-cp310-win_amd64.whl.metadata (2.2 kB)
Collecting cymem<2.1.0,>=2.0.2 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached cymem-2.0.11-cp310-cp310-win_amd64.whl.metadata (8.8 kB)
Collecting preshed<3.1.0,>=3.0.2 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached preshed-3.0.9-cp310-cp310-win_amd64.whl.metadata (2.2 kB)
Collecting thinc<8.4.0,>=8.3.4 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached thinc-8.3.6-cp310-cp310-win_amd64.whl.metadata (15 kB)
Collecting wasabi<1.2.0,>=0.9.1 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached wasabi-1.1.3-py3-none-any.whl.metadata (28 kB)
Collecting srsly<3.0.0,>=2.4.3 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached srsly-2.5.1-cp310-cp310-win_amd64.whl.metadata (20 kB)
Collecting catalogue<2.1.0,>=2.0.6 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached catalogue-2.0.10-py3-none-any.whl.metadata (14 kB)
Collecting weasel<0.5.0,>=0.1.0 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached weasel-0.4.1-py3-none-any.whl.metadata (4.6 kB)
Collecting typer<1.0.0,>=0.3.0 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached typer-0.15.3-py3-none-any.whl.metadata (15 kB)
Requirement already satisfied: setuptools in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon) (75.6.0)
Collecting langcodes<4.0.0,>=3.2.0 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached langcodes-3.5.0-py3-none-any.whl.metadata (29 kB)
Requirement already satisfied: statsmodels>=0.13.2 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from statsforecast<2.0.2,>=1.7.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon) (0.14.4)
Requirement already satisfied: absl-py>=0.4 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from tensorboard<3,>=2.9->autogluon.multimodal==1.3.0->autogluon) (2.1.0)
Requirement already satisfied: markdown>=2.6.8 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from tensorboard<3,>=2.9->autogluon.multimodal==1.3.0->autogluon) (3.7)
Requirement already satisfied: tensorboard-data-server<0.8.0,>=0.7.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from tensorboard<3,>=2.9->autogluon.multimodal==1.3.0->autogluon) (0.7.2)
Requirement already satisfied: werkzeug>=1.0.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from tensorboard<3,>=2.9->autogluon.multimodal==1.3.0->autogluon) (3.1.3)
Collecting sympy==1.13.1 (from torch<2.7,>=2.2->autogluon.multimodal==1.3.0->autogluon)
  Using cached sympy-1.13.1-py3-none-any.whl.metadata (12 kB)
Collecting mpmath<1.4,>=1.1.0 (from sympy==1.13.1->torch<2.7,>=2.2->autogluon.multimodal==1.3.0->autogluon)
  Using cached mpmath-1.3.0-py3-none-any.whl.metadata (8.6 kB)
Collecting tokenizers<0.22,>=0.21 (from transformers<4.50,>=4.38.0->transformers[sentencepiece]<4.50,>=4.38.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached tokenizers-0.21.1-cp39-abi3-win_amd64.whl.metadata (6.9 kB)
Collecting sentencepiece!=0.1.92,>=0.1.91 (from transformers[sentencepiece]<4.50,>=4.38.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached sentencepiece-0.2.0-cp310-cp310-win_amd64.whl.metadata (8.3 kB)
Collecting aiohappyeyeballs>=2.3.0 (from aiohttp>=3.7->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached aiohappyeyeballs-2.6.1-py3-none-any.whl.metadata (5.9 kB)
Collecting async-timeout<6.0,>=4.0 (from aiohttp>=3.7->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached async_timeout-5.0.1-py3-none-any.whl.metadata (5.1 kB)
Collecting multidict<7.0,>=4.5 (from aiohttp>=3.7->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached multidict-6.4.3-cp310-cp310-win_amd64.whl.metadata (5.5 kB)
Collecting propcache>=0.2.0 (from aiohttp>=3.7->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached propcache-0.3.1-cp310-cp310-win_amd64.whl.metadata (11 kB)
Collecting yarl<2.0,>=1.17.0 (from aiohttp>=3.7->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached yarl-1.20.0-cp310-cp310-win_amd64.whl.metadata (74 kB)
Collecting dill (from evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached dill-0.3.8-py3-none-any.whl.metadata (10 kB)
Collecting multiprocess (from evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached multiprocess-0.70.16-py310-none-any.whl.metadata (7.2 kB)
Collecting fsspec>=2021.05.0 (from fsspec[http]>=2021.05.0->evaluate<0.5.0,>=0.4.0->autogluon.multimodal==1.3.0->autogluon)
  Using cached fsspec-2025.3.0-py3-none-any.whl.metadata (11 kB)
Collecting beautifulsoup4 (from gdown>=4.0.0->nlpaug<1.2.0,>=1.1.10->autogluon.multimodal==1.3.0->autogluon)
  Using cached beautifulsoup4-4.13.4-py3-none-any.whl.metadata (3.8 kB)
Collecting language-data>=1.2 (from langcodes<4.0.0,>=3.2.0->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached language_data-1.3.0-py3-none-any.whl.metadata (4.3 kB)
Collecting llvmlite<0.45,>=0.44.0dev0 (from numba->mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached llvmlite-0.44.0-cp310-cp310-win_amd64.whl.metadata (5.0 kB)
Collecting annotated-types>=0.6.0 (from pydantic<3,>=1.7->gluonts<0.17,>=0.15.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached annotated_types-0.7.0-py3-none-any.whl.metadata (15 kB)
Collecting pydantic-core==2.33.2 (from pydantic<3,>=1.7->gluonts<0.17,>=0.15.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached pydantic_core-2.33.2-cp310-cp310-win_amd64.whl.metadata (6.9 kB)
Collecting typing-inspection>=0.4.0 (from pydantic<3,>=1.7->gluonts<0.17,>=0.15.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached typing_inspection-0.4.0-py3-none-any.whl.metadata (2.6 kB)
Requirement already satisfied: patsy>=0.5.6 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from statsmodels>=0.13.2->statsforecast<2.0.2,>=1.7.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon) (1.0.1)
Collecting blis<1.4.0,>=1.3.0 (from thinc<8.4.0,>=8.3.4->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached blis-1.3.0-cp310-cp310-win_amd64.whl.metadata (7.6 kB)
Collecting confection<1.0.0,>=0.0.1 (from thinc<8.4.0,>=8.3.4->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached confection-0.1.5-py3-none-any.whl.metadata (19 kB)
INFO: pip is looking at multiple versions of thinc to determine which version is compatible with other requirements. This could take a while.
Collecting thinc<8.4.0,>=8.3.4 (from spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached thinc-8.3.4-cp310-cp310-win_amd64.whl.metadata (15 kB)
Collecting blis<1.3.0,>=1.2.0 (from thinc<8.4.0,>=8.3.4->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached blis-1.2.1-cp310-cp310-win_amd64.whl.metadata (7.6 kB)
Collecting fs (from triad>=0.9.7->fugue>=0.9.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached fs-2.4.16-py2.py3-none-any.whl.metadata (6.3 kB)
Collecting shellingham>=1.3.0 (from typer<1.0.0,>=0.3.0->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached shellingham-1.5.4-py2.py3-none-any.whl.metadata (3.5 kB)
Requirement already satisfied: markdown-it-py>=2.2.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from rich->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon) (3.0.0)
Requirement already satisfied: pygments<3.0.0,>=2.13.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from rich->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon) (2.18.0)
Collecting distlib<1,>=0.3.7 (from virtualenv!=20.21.1,>=20.0.24->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached distlib-0.3.9-py2.py3-none-any.whl.metadata (5.2 kB)
Requirement already satisfied: platformdirs<5,>=3.9.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from virtualenv!=20.21.1,>=20.0.24->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon) (4.3.6)
Collecting cloudpathlib<1.0.0,>=0.7.0 (from weasel<0.5.0,>=0.1.0->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached cloudpathlib-0.21.0-py3-none-any.whl.metadata (14 kB)
Requirement already satisfied: wrapt in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from smart_open->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon) (1.17.0)
Collecting ordered-set (from model-index->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached ordered_set-4.1.0-py3-none-any.whl.metadata (5.3 kB)
Collecting opencensus-context>=0.1.3 (from opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached opencensus_context-0.1.3-py2.py3-none-any.whl.metadata (3.3 kB)
Collecting google-api-core<3.0.0,>=1.0.0 (from opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached google_api_core-2.24.2-py3-none-any.whl.metadata (3.0 kB)
Collecting pycryptodome (from opendatalab->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached pycryptodome-3.22.0-cp37-abi3-win_amd64.whl.metadata (3.4 kB)
Collecting openxlab (from opendatalab->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached openxlab-0.1.2-py3-none-any.whl.metadata (3.8 kB)
Requirement already satisfied: pywin32 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from opendatalab->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon) (310)
Collecting alembic>=1.5.0 (from optuna->mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached alembic-1.15.2-py3-none-any.whl.metadata (7.3 kB)
Collecting colorlog (from optuna->mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached colorlog-6.9.0-py3-none-any.whl.metadata (10 kB)
Collecting sqlalchemy>=1.4.2 (from optuna->mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached sqlalchemy-2.0.40-cp310-cp310-win_amd64.whl.metadata (9.9 kB)
Requirement already satisfied: tenacity>=6.2.0 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from plotly->catboost<1.3,>=1.2->autogluon.tabular[all]==1.3.0->autogluon) (9.0.0)
Collecting beartype>=0.16.2 (from plum-dispatch->fastai<2.9,>=2.3.1->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached beartype-0.20.2-py3-none-any.whl.metadata (33 kB)
Collecting Mako (from alembic>=1.5.0->optuna->mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached mako-1.3.10-py3-none-any.whl.metadata (2.9 kB)
Collecting googleapis-common-protos<2.0.0,>=1.56.2 (from google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached googleapis_common_protos-1.70.0-py3-none-any.whl.metadata (9.3 kB)
Collecting proto-plus<2.0.0,>=1.22.3 (from google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached proto_plus-1.26.1-py3-none-any.whl.metadata (2.2 kB)
Collecting google-auth<3.0.0,>=2.14.1 (from google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached google_auth-2.40.1-py2.py3-none-any.whl.metadata (6.2 kB)
Collecting marisa-trie>=1.1.0 (from language-data>=1.2->langcodes<4.0.0,>=3.2.0->spacy<3.9->autogluon.tabular[all]==1.3.0->autogluon)
  Using cached marisa_trie-1.2.1-cp310-cp310-win_amd64.whl.metadata (9.3 kB)
Requirement already satisfied: mdurl~=0.1 in c:\users\chunn\anaconda3\envs\ads3\lib\site-packages (from markdown-it-py>=2.2.0->rich->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon) (0.1.2)
Collecting greenlet>=1 (from sqlalchemy>=1.4.2->optuna->mlforecast<0.14,>0.13->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached greenlet-3.2.2-cp310-cp310-win_amd64.whl.metadata (4.2 kB)
Collecting soupsieve>1.2 (from beautifulsoup4->gdown>=4.0.0->nlpaug<1.2.0,>=1.1.10->autogluon.multimodal==1.3.0->autogluon)
  Using cached soupsieve-2.7-py3-none-any.whl.metadata (4.6 kB)
Collecting appdirs~=1.4.3 (from fs->triad>=0.9.7->fugue>=0.9.0->autogluon.timeseries==1.3.0->autogluon.timeseries[all]==1.3.0->autogluon)
  Using cached appdirs-1.4.4-py2.py3-none-any.whl.metadata (9.0 kB)
Collecting filelock (from ray<2.45,>=2.10.0->ray[default]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached filelock-3.14.0-py3-none-any.whl.metadata (2.8 kB)
Collecting oss2~=2.17.0 (from openxlab->opendatalab->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached oss2-2.17.0.tar.gz (259 kB)
  Preparing metadata (setup.py): started
  Preparing metadata (setup.py): finished with status 'done'
Collecting pytz>=2020.1 (from pandas<2.3.0,>=2.0.0->autogluon.core==1.3.0->autogluon.core[all]==1.3.0->autogluon)
  Using cached pytz-2023.4-py2.py3-none-any.whl.metadata (22 kB)
INFO: pip is looking at multiple versions of openxlab to determine which version is compatible with other requirements. This could take a while.
Collecting openxlab (from opendatalab->openmim<0.4.0,>=0.3.7->autogluon.multimodal==1.3.0->autogluon)
  Using cached openxlab-0.1.1-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.1.0-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.38-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.37-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.36-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.35-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.34-py3-none-any.whl.metadata (3.8 kB)
INFO: pip is still looking at multiple versions of openxlab to determine which version is compatible with other requirements. This could take a while.
  Using cached openxlab-0.0.33-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.32-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.31-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.30-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.29-py3-none-any.whl.metadata (3.8 kB)
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. See https://pip.pypa.io/warnings/backtracking for guidance. If you want to abort this run, press Ctrl + C.
  Using cached openxlab-0.0.28-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.27-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.26-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.25-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.24-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.23-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.22-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.21-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.20-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.19-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.18-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.17-py3-none-any.whl.metadata (3.7 kB)
  Using cached openxlab-0.0.16-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.15-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.14-py3-none-any.whl.metadata (3.8 kB)
  Using cached openxlab-0.0.13-py3-none-any.whl.metadata (4.5 kB)
  Using cached openxlab-0.0.12-py3-none-any.whl.metadata (4.5 kB)
  Using cached openxlab-0.0.11-py3-none-any.whl.metadata (4.3 kB)
Collecting PySocks!=1.5.7,>=1.5.6 (from requests[socks]->gdown>=4.0.0->nlpaug<1.2.0,>=1.1.10->autogluon.multimodal==1.3.0->autogluon)
  Using cached PySocks-1.7.1-py3-none-any.whl.metadata (13 kB)
Collecting cachetools<6.0,>=2.0.0 (from google-auth<3.0.0,>=2.14.1->google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached cachetools-5.5.2-py3-none-any.whl.metadata (5.4 kB)
Collecting pyasn1-modules>=0.2.1 (from google-auth<3.0.0,>=2.14.1->google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached pyasn1_modules-0.4.2-py3-none-any.whl.metadata (3.5 kB)
Collecting rsa<5,>=3.1.4 (from google-auth<3.0.0,>=2.14.1->google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached rsa-4.9.1-py3-none-any.whl.metadata (5.6 kB)
Collecting pyasn1<0.7.0,>=0.6.1 (from pyasn1-modules>=0.2.1->google-auth<3.0.0,>=2.14.1->google-api-core<3.0.0,>=1.0.0->opencensus->ray[default,tune]<2.45,>=2.10.0; extra == "all"->autogluon.core[all]==1.3.0->autogluon)
  Using cached pyasn1-0.6.1-py3-none-any.whl.metadata (8.4 kB)
Using cached autogluon-1.3.0-py3-none-any.whl (9.8 kB)
Using cached autogluon.core-1.3.0-py3-none-any.whl (222 kB)
Using cached autogluon.features-1.3.0-py3-none-any.whl (64 kB)
Using cached autogluon.multimodal-1.3.0-py3-none-any.whl (454 kB)
Using cached autogluon.tabular-1.3.0-py3-none-any.whl (382 kB)
Using cached autogluon.timeseries-1.3.0-py3-none-any.whl (181 kB)
Using cached autogluon.common-1.3.0-py3-none-any.whl (69 kB)
Using cached accelerate-1.6.0-py3-none-any.whl (354 kB)
Using cached boto3-1.38.13-py3-none-any.whl (139 kB)
Using cached coreforecast-0.0.15-cp310-cp310-win_amd64.whl (189 kB)
Using cached defusedxml-0.7.1-py2.py3-none-any.whl (25 kB)
Using cached einops-0.8.1-py3-none-any.whl (64 kB)
Using cached evaluate-0.4.3-py3-none-any.whl (84 kB)
Using cached fastai-2.8.1-py3-none-any.whl (235 kB)
Using cached fugue-0.9.1-py3-none-any.whl (278 kB)
Using cached gluonts-0.16.1-py3-none-any.whl (1.5 MB)
Using cached hyperopt-0.2.7-py2.py3-none-any.whl (1.6 MB)
Using cached jsonschema-4.23.0-py3-none-any.whl (88 kB)
Using cached lightning-2.5.1.post0-py3-none-any.whl (819 kB)
Using cached mlforecast-0.13.6-py3-none-any.whl (71 kB)
Using cached networkx-3.4.2-py3-none-any.whl (1.7 MB)
Using cached nlpaug-1.1.11-py3-none-any.whl (410 kB)
Using cached nltk-3.9.1-py3-none-any.whl (1.5 MB)
Using cached omegaconf-2.3.0-py3-none-any.whl (79 kB)
Using cached openmim-0.3.9-py2.py3-none-any.whl (52 kB)
Using cached orjson-3.10.18-cp310-cp310-win_amd64.whl (134 kB)
Using cached pdf2image-1.17.0-py3-none-any.whl (11 kB)
Using cached pyarrow-20.0.0-cp310-cp310-win_amd64.whl (25.8 MB)
Using cached pytesseract-0.3.13-py3-none-any.whl (14 kB)
Using cached pytorch_metric_learning-2.8.1-py3-none-any.whl (125 kB)
Using cached ray-2.44.1-cp310-cp310-win_amd64.whl (25.7 MB)
Using cached scikit_image-0.25.2-cp310-cp310-win_amd64.whl (12.8 MB)
Using cached spacy-3.8.5-cp310-cp310-win_amd64.whl (12.2 MB)
Using cached statsforecast-2.0.1-cp310-cp310-win_amd64.whl (275 kB)
Using cached text_unidecode-1.3-py2.py3-none-any.whl (78 kB)
Using cached timm-1.0.3-py3-none-any.whl (2.3 MB)
Using cached torch-2.6.0-cp310-cp310-win_amd64.whl (204.2 MB)
Using cached sympy-1.13.1-py3-none-any.whl (6.2 MB)
Using cached torchmetrics-1.7.1-py3-none-any.whl (961 kB)
Using cached torchvision-0.21.0-cp310-cp310-win_amd64.whl (1.6 MB)
Using cached tqdm-4.67.1-py3-none-any.whl (78 kB)
Using cached transformers-4.49.0-py3-none-any.whl (10.0 MB)
Using cached utilsforecast-0.2.10-py3-none-any.whl (41 kB)
Using cached pytorch_lightning-2.5.1.post0-py3-none-any.whl (823 kB)
Using cached adagio-0.2.6-py3-none-any.whl (19 kB)
Using cached aiohttp-3.11.18-cp310-cp310-win_amd64.whl (442 kB)
Using cached aiosignal-1.3.2-py2.py3-none-any.whl (7.6 kB)
Using cached attrs-25.3.0-py3-none-any.whl (63 kB)
Using cached botocore-1.38.13-py3-none-any.whl (13.5 MB)
Using cached catalogue-2.0.10-py3-none-any.whl (17 kB)
Downloading click-8.2.0-py3-none-any.whl (102 kB)
Using cached cymem-2.0.11-cp310-cp310-win_amd64.whl (39 kB)
Using cached datasets-3.6.0-py3-none-any.whl (491 kB)
Using cached dill-0.3.8-py3-none-any.whl (116 kB)
Using cached fastcore-1.8.2-py3-none-any.whl (78 kB)
Using cached fastdownload-0.0.7-py3-none-any.whl (12 kB)
Using cached fastprogress-1.0.3-py3-none-any.whl (12 kB)
Using cached fasttransform-0.0.2-py3-none-any.whl (14 kB)
Using cached frozenlist-1.6.0-cp310-cp310-win_amd64.whl (120 kB)
Using cached fsspec-2025.3.0-py3-none-any.whl (193 kB)
Using cached gdown-5.2.0-py3-none-any.whl (18 kB)
Using cached huggingface_hub-0.31.1-py3-none-any.whl (484 kB)
Using cached imageio-2.37.0-py3-none-any.whl (315 kB)
Using cached jmespath-1.0.1-py3-none-any.whl (20 kB)
Using cached jsonschema_specifications-2025.4.1-py3-none-any.whl (18 kB)
Using cached langcodes-3.5.0-py3-none-any.whl (182 kB)
Using cached lazy_loader-0.4-py3-none-any.whl (12 kB)
Using cached lightning_utilities-0.14.3-py3-none-any.whl (28 kB)
Using cached msgpack-1.1.0-cp310-cp310-win_amd64.whl (74 kB)
Using cached multiprocess-0.70.16-py310-none-any.whl (134 kB)
Using cached murmurhash-1.0.12-cp310-cp310-win_amd64.whl (25 kB)
Using cached numba-0.61.2-cp310-cp310-win_amd64.whl (2.8 MB)
Using cached preshed-3.0.9-cp310-cp310-win_amd64.whl (122 kB)
Using cached prometheus_client-0.21.1-py3-none-any.whl (54 kB)
Using cached py_spy-0.4.0-py2.py3-none-win_amd64.whl (1.8 MB)
Using cached pydantic-2.11.4-py3-none-any.whl (443 kB)
Using cached pydantic_core-2.33.2-cp310-cp310-win_amd64.whl (2.0 MB)
Using cached PyYAML-6.0.2-cp310-cp310-win_amd64.whl (161 kB)
Using cached referencing-0.36.2-py3-none-any.whl (26 kB)
Using cached regex-2024.11.6-cp310-cp310-win_amd64.whl (274 kB)
Using cached rpds_py-0.24.0-cp310-cp310-win_amd64.whl (234 kB)
Using cached s3transfer-0.12.0-py3-none-any.whl (84 kB)
Using cached safetensors-0.5.3-cp38-abi3-win_amd64.whl (308 kB)
Using cached sentencepiece-0.2.0-cp310-cp310-win_amd64.whl (991 kB)
Using cached spacy_legacy-3.0.12-py2.py3-none-any.whl (29 kB)
Using cached spacy_loggers-1.0.5-py3-none-any.whl (22 kB)
Using cached srsly-2.5.1-cp310-cp310-win_amd64.whl (632 kB)
Using cached tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB)
Using cached thinc-8.3.4-cp310-cp310-win_amd64.whl (1.5 MB)
Using cached tifffile-2025.5.10-py3-none-any.whl (226 kB)
Using cached tokenizers-0.21.1-cp39-abi3-win_amd64.whl (2.4 MB)
Using cached toolz-0.12.1-py3-none-any.whl (56 kB)
Using cached triad-0.9.8-py3-none-any.whl (62 kB)
Using cached typer-0.15.3-py3-none-any.whl (45 kB)
Using cached virtualenv-20.31.2-py3-none-any.whl (6.1 MB)
Using cached filelock-3.18.0-py3-none-any.whl (16 kB)
Using cached wasabi-1.1.3-py3-none-any.whl (27 kB)
Using cached weasel-0.4.1-py3-none-any.whl (50 kB)
Using cached smart_open-7.1.0-py3-none-any.whl (61 kB)
Using cached aiohttp_cors-0.8.1-py3-none-any.whl (25 kB)
Using cached cloudpickle-3.1.1-py3-none-any.whl (20 kB)
Using cached colorful-0.5.6-py2.py3-none-any.whl (201 kB)
Using cached future-1.0.0-py3-none-any.whl (491 kB)
Using cached model_index-0.1.11-py3-none-any.whl (34 kB)
Using cached opencensus-0.11.4-py2.py3-none-any.whl (128 kB)
Using cached opendatalab-0.0.10-py3-none-any.whl (29 kB)
Using cached optuna-4.3.0-py3-none-any.whl (386 kB)
Using cached plum_dispatch-2.5.7-py3-none-any.whl (42 kB)
Using cached py4j-0.10.9.9-py2.py3-none-any.whl (203 kB)
Using cached tabulate-0.9.0-py3-none-any.whl (35 kB)
Using cached window_ops-0.0.15-py3-none-any.whl (15 kB)
Using cached xxhash-3.5.0-cp310-cp310-win_amd64.whl (30 kB)
Using cached aiohappyeyeballs-2.6.1-py3-none-any.whl (15 kB)
Using cached alembic-1.15.2-py3-none-any.whl (231 kB)
Using cached annotated_types-0.7.0-py3-none-any.whl (13 kB)
Using cached async_timeout-5.0.1-py3-none-any.whl (6.2 kB)
Using cached beartype-0.20.2-py3-none-any.whl (1.2 MB)
Using cached blis-1.2.1-cp310-cp310-win_amd64.whl (6.2 MB)
Using cached cloudpathlib-0.21.0-py3-none-any.whl (52 kB)
Using cached confection-0.1.5-py3-none-any.whl (35 kB)
Using cached distlib-0.3.9-py2.py3-none-any.whl (468 kB)
Using cached google_api_core-2.24.2-py3-none-any.whl (160 kB)
Using cached language_data-1.3.0-py3-none-any.whl (5.4 MB)
Using cached llvmlite-0.44.0-cp310-cp310-win_amd64.whl (30.3 MB)
Using cached mpmath-1.3.0-py3-none-any.whl (536 kB)
Using cached multidict-6.4.3-cp310-cp310-win_amd64.whl (38 kB)
Using cached opencensus_context-0.1.3-py2.py3-none-any.whl (5.1 kB)
Using cached propcache-0.3.1-cp310-cp310-win_amd64.whl (45 kB)
Using cached shellingham-1.5.4-py2.py3-none-any.whl (9.8 kB)
Using cached sqlalchemy-2.0.40-cp310-cp310-win_amd64.whl (2.1 MB)
Using cached typing_inspection-0.4.0-py3-none-any.whl (14 kB)
Using cached yarl-1.20.0-cp310-cp310-win_amd64.whl (92 kB)
Using cached beautifulsoup4-4.13.4-py3-none-any.whl (187 kB)
Using cached colorlog-6.9.0-py3-none-any.whl (11 kB)
Using cached fs-2.4.16-py2.py3-none-any.whl (135 kB)
Using cached openxlab-0.0.11-py3-none-any.whl (55 kB)
Using cached ordered_set-4.1.0-py3-none-any.whl (7.6 kB)
Using cached pycryptodome-3.22.0-cp37-abi3-win_amd64.whl (1.8 MB)
Using cached appdirs-1.4.4-py2.py3-none-any.whl (9.6 kB)
Using cached google_auth-2.40.1-py2.py3-none-any.whl (216 kB)
Using cached googleapis_common_protos-1.70.0-py3-none-any.whl (294 kB)
Using cached greenlet-3.2.2-cp310-cp310-win_amd64.whl (294 kB)
Using cached marisa_trie-1.2.1-cp310-cp310-win_amd64.whl (151 kB)
Using cached proto_plus-1.26.1-py3-none-any.whl (50 kB)
Using cached PySocks-1.7.1-py3-none-any.whl (16 kB)
Using cached soupsieve-2.7-py3-none-any.whl (36 kB)
Using cached mako-1.3.10-py3-none-any.whl (78 kB)
Using cached cachetools-5.5.2-py3-none-any.whl (10 kB)
Using cached pyasn1_modules-0.4.2-py3-none-any.whl (181 kB)
Using cached rsa-4.9.1-py3-none-any.whl (34 kB)
Using cached pyasn1-0.6.1-py3-none-any.whl (83 kB)
Installing collected packages: text-unidecode, sentencepiece, py4j, py-spy, opencensus-context, nvidia-ml-py3, mpmath, distlib, cymem, appdirs, antlr4-python3-runtime, xxhash, wasabi, typing-inspection, tqdm, toolz, tifffile, tensorboardX, tabulate, sympy, spacy-loggers, spacy-legacy, soupsieve, smart_open, shellingham, safetensors, rpds-py, regex, pyyaml, pytesseract, PySocks, pydantic-core, pycryptodome, pyasn1, pyarrow, proto-plus, propcache, prometheus_client, pdf2image, orjson, ordered-set, openxlab, networkx, murmurhash, multidict, msgpack, marisa-trie, Mako, llvmlite, lightning-utilities, lazy-loader, jmespath, imageio, greenlet, googleapis-common-protos, future, fsspec, fs, frozenlist, filelock, fastprogress, fastcore, einops, dill, defusedxml, coreforecast, colorlog, colorful, cloudpickle, cloudpathlib, click, catalogue, cachetools, blis, beartype, attrs, async-timeout, annotated-types, aiohappyeyeballs, yarl, virtualenv, torch, srsly, sqlalchemy, scikit-image, rsa, referencing, pydantic, pyasn1-modules, preshed, omegaconf, numba, nltk, multiprocess, model-index, language-data, hyperopt, huggingface-hub, fastdownload, botocore, beautifulsoup4, aiosignal, window-ops, utilsforecast, typer, triad, torchvision, torchmetrics, tokenizers, seqeval, s3transfer, pytorch-metric-learning, plum-dispatch, opendatalab, langcodes, jsonschema-specifications, google-auth, gluonts, gdown, confection, alembic, aiohttp, accelerate, weasel, transformers, timm, thinc, optuna, openmim, nlpaug, jsonschema, google-api-core, fasttransform, boto3, aiohttp_cors, adagio, spacy, ray, pytorch-lightning, opencensus, mlforecast, fugue, datasets, autogluon.common, statsforecast, lightning, fastai, evaluate, autogluon.features, autogluon.core, autogluon.tabular, autogluon.multimodal, autogluon.timeseries, autogluon
Successfully installed Mako-1.3.10 PySocks-1.7.1 accelerate-1.6.0 adagio-0.2.6 aiohappyeyeballs-2.6.1 aiohttp-3.11.18 aiohttp_cors-0.8.1 aiosignal-1.3.2 alembic-1.15.2 annotated-types-0.7.0 antlr4-python3-runtime-4.9.3 appdirs-1.4.4 async-timeout-5.0.1 attrs-25.3.0 autogluon-1.3.0 autogluon.common-1.3.0 autogluon.core-1.3.0 autogluon.features-1.3.0 autogluon.multimodal-1.3.0 autogluon.tabular-1.3.0 autogluon.timeseries-1.3.0 beartype-0.20.2 beautifulsoup4-4.13.4 blis-1.2.1 boto3-1.38.13 botocore-1.38.13 cachetools-5.5.2 catalogue-2.0.10 click-8.2.0 cloudpathlib-0.21.0 cloudpickle-3.1.1 colorful-0.5.6 colorlog-6.9.0 confection-0.1.5 coreforecast-0.0.15 cymem-2.0.11 datasets-3.6.0 defusedxml-0.7.1 dill-0.3.8 distlib-0.3.9 einops-0.8.1 evaluate-0.4.3 fastai-2.8.1 fastcore-1.8.2 fastdownload-0.0.7 fastprogress-1.0.3 fasttransform-0.0.2 filelock-3.18.0 frozenlist-1.6.0 fs-2.4.16 fsspec-2025.3.0 fugue-0.9.1 future-1.0.0 gdown-5.2.0 gluonts-0.16.1 google-api-core-2.24.2 google-auth-2.40.1 googleapis-common-protos-1.70.0 greenlet-3.2.2 huggingface-hub-0.31.1 hyperopt-0.2.7 imageio-2.37.0 jmespath-1.0.1 jsonschema-4.23.0 jsonschema-specifications-2025.4.1 langcodes-3.5.0 language-data-1.3.0 lazy-loader-0.4 lightning-2.5.1.post0 lightning-utilities-0.14.3 llvmlite-0.44.0 marisa-trie-1.2.1 mlforecast-0.13.6 model-index-0.1.11 mpmath-1.3.0 msgpack-1.1.0 multidict-6.4.3 multiprocess-0.70.16 murmurhash-1.0.12 networkx-3.4.2 nlpaug-1.1.11 nltk-3.9.1 numba-0.61.2 nvidia-ml-py3-7.352.0 omegaconf-2.3.0 opencensus-0.11.4 opencensus-context-0.1.3 opendatalab-0.0.10 openmim-0.3.9 openxlab-0.0.11 optuna-4.3.0 ordered-set-4.1.0 orjson-3.10.18 pdf2image-1.17.0 plum-dispatch-2.5.7 preshed-3.0.9 prometheus_client-0.21.1 propcache-0.3.1 proto-plus-1.26.1 py-spy-0.4.0 py4j-0.10.9.9 pyarrow-20.0.0 pyasn1-0.6.1 pyasn1-modules-0.4.2 pycryptodome-3.22.0 pydantic-2.11.4 pydantic-core-2.33.2 pytesseract-0.3.13 pytorch-lightning-2.5.1.post0 pytorch-metric-learning-2.8.1 pyyaml-6.0.2 ray-2.44.1 referencing-0.36.2 regex-2024.11.6 rpds-py-0.24.0 rsa-4.9.1 s3transfer-0.12.0 safetensors-0.5.3 scikit-image-0.25.2 sentencepiece-0.2.0 seqeval-1.2.2 shellingham-1.5.4 smart_open-7.1.0 soupsieve-2.7 spacy-3.8.5 spacy-legacy-3.0.12 spacy-loggers-1.0.5 sqlalchemy-2.0.40 srsly-2.5.1 statsforecast-2.0.1 sympy-1.13.1 tabulate-0.9.0 tensorboardX-2.6.2.2 text-unidecode-1.3 thinc-8.3.4 tifffile-2025.5.10 timm-1.0.3 tokenizers-0.21.1 toolz-0.12.1 torch-2.6.0 torchmetrics-1.7.1 torchvision-0.21.0 tqdm-4.67.1 transformers-4.49.0 triad-0.9.8 typer-0.15.3 typing-inspection-0.4.0 utilsforecast-0.2.10 virtualenv-20.31.2 wasabi-1.1.3 weasel-0.4.1 window-ops-0.0.15 xxhash-3.5.0 yarl-1.20.0

[notice] A new release of pip is available: 24.3.1 -> 25.1.1
[notice] To update, run: python.exe -m pip install --upgrade pip
c:\Users\chunn\anaconda3\envs\ads3\lib\site-packages\tqdm\auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

# Combine features and target variable into a single dataframe as reuqired by AutoGluon
Xy_train = X_train.copy()
Xy_val = X_val.copy()

Xy_train['TARGET'] = y_train 
Xy_val['TARGET'] = y_val.reset_index(drop=True) # Since StandardScaler above has reset the index

# Start timer to calculate the running time of training the CatBoost model
start_time = time.time()

# Ensure reproducibility
set_random_seed(RANDOM_SEED)

# Initialize and fit the AutoGluon predictor
AG = TabularPredictor(label='TARGET', eval_metric='roc_auc', path='autogluon_diabetes_auc').fit(
    train_data=Xy_train,
    time_limit=3600,         # One hour to search for the best model
    presets='best_quality',  # Enables stacking, bagging, hyperparameter tuning for best possible performance
    auto_stack=True,         # Explicit stacking (included in 'best_quality')
    num_stack_levels=2       # Number of meta-model layers
)

elapsed_time_AG = time.time() - start_time
print(f"Elapsed time (sec): {elapsed_time_AG:.2f}")

Verbosity: 2 (Standard Logging)
=================== System Info ===================
AutoGluon Version:  1.3.0
Python Version:     3.10.16
Operating System:   Windows
Platform Machine:   AMD64
Platform Version:   10.0.26100
CPU Count:          12
Memory Avail:       2.35 GB / 15.73 GB (14.9%)
Disk Space Avail:   512.29 GB / 936.72 GB (54.7%)
===================================================
Presets specified: ['best_quality']
Setting dynamic_stacking from 'auto' to True. Reason: Enable dynamic_stacking when use_bag_holdout is disabled. (use_bag_holdout=False)
Stack configuration (auto_stack=True): num_stack_levels=2, num_bag_folds=8, num_bag_sets=1
DyStack is enabled (dynamic_stacking=True). AutoGluon will try to determine whether the input data is affected by stacked overfitting and enable or disable stacking as a consequence.
	This is used to identify the optimal `num_stack_levels` value. Copies of AutoGluon will be fit on subsets of the data. Then holdout validation data is used to detect stacked overfitting.
	Running DyStack for up to 900s of the 3600s of remaining time (25%).
2025-05-13 11:24:07,066	INFO util.py:154 -- Missing packages: ['ipywidgets']. Run `pip install -U ipywidgets`, then restart the notebook server for rich notebook output.
	Running DyStack sub-fit in a ray process to avoid memory leakage. Enabling ray logging (enable_ray_logging=True). Specify `ds_args={'enable_ray_logging': False}` if you experience logging issues.
2025-05-13 11:24:12,436	INFO worker.py:1843 -- Started a local Ray instance. View the dashboard at 127.0.0.1:8266 
		Context path: "c:\studium\DAV Ausbildung\CADS\Immersion\ads3_exam\03_scripts\autogluon_diabetes_auc\ds_sub_fit\sub_fit_ho"
(_dystack pid=32908) Running DyStack sub-fit ...
(_dystack pid=32908) Beginning AutoGluon training ... Time limit = 892s
(_dystack pid=32908) AutoGluon will save models to "c:\studium\DAV Ausbildung\CADS\Immersion\ads3_exam\03_scripts\autogluon_diabetes_auc\ds_sub_fit\sub_fit_ho"
(_dystack pid=32908) Train Data Rows:    11712
(_dystack pid=32908) Train Data Columns: 2248
(_dystack pid=32908) Label Column:       TARGET
(_dystack pid=32908) Problem Type:       binary
(_dystack pid=32908) Preprocessing data ...
(_dystack pid=32908) Selected class <--> label mapping:  class 1 = 1, class 0 = 0
(_dystack pid=32908) Using Feature Generators to preprocess the data ...
(_dystack pid=32908) Fitting AutoMLPipelineFeatureGenerator...
(_dystack pid=32908) 	Available Memory:                    2322.95 MB
(_dystack pid=32908) 	Train Data (Original)  Memory Usage: 200.87 MB (8.6% of available memory)
(_dystack pid=32908) 	Warning: Data size prior to feature transformation consumes 8.6% of available memory. Consider increasing memory or subsampling the data to avoid instability.
(_dystack pid=32908) 	Inferring data type of each feature based on column values. Set feature_metadata_in to manually specify special dtypes of the features.
(_dystack pid=32908) 	Stage 1 Generators:
(_dystack pid=32908) 		Fitting AsTypeFeatureGenerator...
(_dystack pid=32908) 			Note: Converting 1667 features to boolean dtype as they only contain 2 unique values.
(_dystack pid=32908) 	Stage 2 Generators:
(_dystack pid=32908) 		Fitting FillNaFeatureGenerator...
(_dystack pid=32908) 	Stage 3 Generators:
(_dystack pid=32908) 		Fitting IdentityFeatureGenerator...
(_dystack pid=32908) 	Stage 4 Generators:
(_dystack pid=32908) 		Fitting DropUniqueFeatureGenerator...
(_dystack pid=32908) 	Stage 5 Generators:
(_dystack pid=32908) 		Fitting DropDuplicatesFeatureGenerator...
(_dystack pid=32908) 	Useless Original Features (Count: 573): ['diag_1_110', 'diag_1_143', 'diag_1_146', 'diag_1_147', 'diag_1_149', 'diag_1_160', 'diag_1_170', 'diag_1_173', 'diag_1_175', 'diag_1_187', 'diag_1_192', 'diag_1_194', 'diag_1_195', 'diag_1_200', 'diag_1_201', 'diag_1_207', 'diag_1_208', 'diag_1_210', 'diag_1_212', 'diag_1_216', 'diag_1_219', 'diag_1_228', 'diag_1_236', 'diag_1_246', 'diag_1_25021', 'diag_1_25032', 'diag_1_2505', 'diag_1_25053', 'diag_1_2509', 'diag_1_25091', 'diag_1_262', 'diag_1_266', 'diag_1_272', 'diag_1_301', 'diag_1_308', 'diag_1_318', 'diag_1_320', 'diag_1_322', 'diag_1_334', 'diag_1_335', 'diag_1_336', 'diag_1_337', 'diag_1_353', 'diag_1_36', 'diag_1_362', 'diag_1_363', 'diag_1_374', 'diag_1_377', 'diag_1_378', 'diag_1_379', 'diag_1_381', 'diag_1_382', 'diag_1_383', 'diag_1_385', 'diag_1_388', 'diag_1_39', 'diag_1_397', 'diag_1_412', 'diag_1_422', 'diag_1_445', 'diag_1_448', 'diag_1_454', 'diag_1_457', 'diag_1_470', 'diag_1_477', 'diag_1_48', 'diag_1_49', 'diag_1_495', 'diag_1_501', 'diag_1_52', 'diag_1_521', 'diag_1_524', 'diag_1_526', 'diag_1_527', 'diag_1_528', 'diag_1_529', 'diag_1_542', 'diag_1_551', 'diag_1_565', 'diag_1_57', 'diag_1_573', 'diag_1_580', 'diag_1_582', 'diag_1_588', 'diag_1_603', 'diag_1_605', 'diag_1_61', 'diag_1_610', 'diag_1_616', 'diag_1_632', 'diag_1_634', 'diag_1_643', 'diag_1_645', 'diag_1_646', 'diag_1_649', 'diag_1_653', 'diag_1_657', 'diag_1_66', 'diag_1_665', 'diag_1_669', 'diag_1_671', 'diag_1_683', 'diag_1_684', 'diag_1_685', 'diag_1_690', 'diag_1_691', 'diag_1_692', 'diag_1_693', 'diag_1_695', 'diag_1_7', 'diag_1_706', 'diag_1_709', 'diag_1_732', 'diag_1_734', 'diag_1_735', 'diag_1_745', 'diag_1_746', 'diag_1_747', 'diag_1_75', 'diag_1_791', 'diag_1_792', 'diag_1_795', 'diag_1_800', 'diag_1_810', 'diag_1_817', 'diag_1_831', 'diag_1_832', 'diag_1_835', 'diag_1_836', 'diag_1_84', 'diag_1_840', 'diag_1_845', 'diag_1_846', 'diag_1_848', 'diag_1_854', 'diag_1_862', 'diag_1_863', 'diag_1_868', 'diag_1_870', 'diag_1_871', 'diag_1_875', 'diag_1_879', 'diag_1_885', 'diag_1_886', 'diag_1_891', 'diag_1_892', 'diag_1_893', 'diag_1_897', 'diag_1_903', 'diag_1_906', 'diag_1_911', 'diag_1_914', 'diag_1_916', 'diag_1_921', 'diag_1_934', 'diag_1_936', 'diag_1_941', 'diag_1_955', 'diag_1_957', 'diag_1_963', 'diag_1_968', 'diag_1_976', 'diag_1_977', 'diag_1_98', 'diag_1_982', 'diag_1_983', 'diag_1_987', 'diag_1_994', 'diag_1_E909', 'diag_1_V45', 'diag_1_V51', 'diag_1_V55', 'diag_1_V56', 'diag_1_V70', 'diag_2_111', 'diag_2_114', 'diag_2_123', 'diag_2_131', 'diag_2_137', 'diag_2_145', 'diag_2_152', 'diag_2_163', 'diag_2_164', 'diag_2_173', 'diag_2_180', 'diag_2_188', 'diag_2_192', 'diag_2_212', 'diag_2_217', 'diag_2_225', 'diag_2_228', 'diag_2_235', 'diag_2_239', 'diag_2_240', 'diag_2_25023', 'diag_2_259', 'diag_2_268', 'diag_2_269', 'diag_2_27', 'diag_2_271', 'diag_2_281', 'diag_2_297', 'diag_2_306', 'diag_2_308', 'diag_2_31', 'diag_2_310', 'diag_2_316', 'diag_2_317', 'diag_2_324', 'diag_2_333', 'diag_2_338', 'diag_2_34', 'diag_2_347', 'diag_2_35', 'diag_2_351', 'diag_2_352', 'diag_2_353', 'diag_2_354', 'diag_2_355', 'diag_2_369', 'diag_2_373', 'diag_2_374', 'diag_2_376', 'diag_2_379', 'diag_2_382', 'diag_2_383', 'diag_2_389', 'diag_2_394', 'diag_2_395', 'diag_2_40', 'diag_2_422', 'diag_2_430', 'diag_2_442', 'diag_2_464', 'diag_2_470', 'diag_2_484', 'diag_2_501', 'diag_2_514', 'diag_2_520', 'diag_2_521', 'diag_2_522', 'diag_2_523', 'diag_2_527', 'diag_2_529', 'diag_2_534', 'diag_2_550', 'diag_2_557', 'diag_2_570', 'diag_2_579', 'diag_2_580', 'diag_2_588', 'diag_2_598', 'diag_2_604', 'diag_2_605', 'diag_2_615', 'diag_2_619', 'diag_2_622', 'diag_2_623', 'diag_2_627', 'diag_2_634', 'diag_2_641', 'diag_2_644', 'diag_2_645', 'diag_2_652', 'diag_2_658', 'diag_2_659', 'diag_2_661', 'diag_2_663', 'diag_2_664', 'diag_2_665', 'diag_2_674', 'diag_2_683', 'diag_2_686', 'diag_2_696', 'diag_2_702', 'diag_2_703', 'diag_2_704', 'diag_2_705', 'diag_2_706', 'diag_2_709', 'diag_2_713', 'diag_2_725', 'diag_2_734', 'diag_2_741', 'diag_2_742', 'diag_2_747', 'diag_2_748', 'diag_2_751', 'diag_2_753', 'diag_2_756', 'diag_2_758', 'diag_2_759', 'diag_2_795', 'diag_2_797', 'diag_2_800', 'diag_2_801', 'diag_2_814', 'diag_2_815', 'diag_2_831', 'diag_2_832', 'diag_2_837', 'diag_2_843', 'diag_2_847', 'diag_2_861', 'diag_2_863', 'diag_2_867', 'diag_2_868', 'diag_2_870', 'diag_2_872', 'diag_2_880', 'diag_2_906', 'diag_2_910', 'diag_2_913', 'diag_2_915', 'diag_2_918', 'diag_2_922', 'diag_2_927', 'diag_2_933', 'diag_2_942', 'diag_2_944', 'diag_2_947', 'diag_2_952', 'diag_2_953', 'diag_2_962', 'diag_2_965', 'diag_2_967', 'diag_2_969', 'diag_2_99', 'diag_2_991', 'diag_2_992', 'diag_2_E816', 'diag_2_E819', 'diag_2_E821', 'diag_2_E850', 'diag_2_E868', 'diag_2_E882', 'diag_2_E883', 'diag_2_E887', 'diag_2_E890', 'diag_2_E905', 'diag_2_E906', 'diag_2_E916', 'diag_2_E924', 'diag_2_E928', 'diag_2_E930', 'diag_2_E931', 'diag_2_E936', 'diag_2_E938', 'diag_2_E944', 'diag_2_E965', 'diag_2_E968', 'diag_2_E980', 'diag_2_V11', 'diag_2_V13', 'diag_2_V16', 'diag_2_V23', 'diag_2_V25', 'diag_2_V46', 'diag_2_V50', 'diag_2_V53', 'diag_2_V57', 'diag_2_V60', 'diag_2_V61', 'diag_2_V86', 'diag_3_115', 'diag_3_117', 'diag_3_122', 'diag_3_123', 'diag_3_136', 'diag_3_139', 'diag_3_14', 'diag_3_141', 'diag_3_150', 'diag_3_152', 'diag_3_161', 'diag_3_163', 'diag_3_17', 'diag_3_170', 'diag_3_171', 'diag_3_172', 'diag_3_182', 'diag_3_193', 'diag_3_214', 'diag_3_217', 'diag_3_223', 'diag_3_227', 'diag_3_233', 'diag_3_235', 'diag_3_240', 'diag_3_245', 'diag_3_25022', 'diag_3_2503', 'diag_3_25031', 'diag_3_258', 'diag_3_259', 'diag_3_265', 'diag_3_270', 'diag_3_271', 'diag_3_299', 'diag_3_314', 'diag_3_315', 'diag_3_318', 'diag_3_323', 'diag_3_35', 'diag_3_350', 'diag_3_351', 'diag_3_359', 'diag_3_360', 'diag_3_361', 'diag_3_36544', 'diag_3_366', 'diag_3_370', 'diag_3_373', 'diag_3_374', 'diag_3_376', 'diag_3_379', 'diag_3_380', 'diag_3_381', 'diag_3_382', 'diag_3_383', 'diag_3_385', 'diag_3_387', 'diag_3_388', 'diag_3_395', 'diag_3_405', 'diag_3_417', 'diag_3_421', 'diag_3_430', 'diag_3_432', 'diag_3_436', 'diag_3_460', 'diag_3_463', 'diag_3_464', 'diag_3_47', 'diag_3_470', 'diag_3_472', 'diag_3_480', 'diag_3_49', 'diag_3_5', 'diag_3_500', 'diag_3_501', 'diag_3_506', 'diag_3_508', 'diag_3_514', 'diag_3_516', 'diag_3_522', 'diag_3_523', 'diag_3_524', 'diag_3_525', 'diag_3_527', 'diag_3_534', 'diag_3_540', 'diag_3_542', 'diag_3_543', 'diag_3_556', 'diag_3_565', 'diag_3_566', 'diag_3_579', 'diag_3_597', 'diag_3_598', 'diag_3_602', 'diag_3_608', 'diag_3_624', 'diag_3_641', 'diag_3_647', 'diag_3_655', 'diag_3_657', 'diag_3_665', 'diag_3_669', 'diag_3_674', 'diag_3_685', 'diag_3_690', 'diag_3_694', 'diag_3_702', 'diag_3_706', 'diag_3_712', 'diag_3_734', 'diag_3_735', 'diag_3_741', 'diag_3_744', 'diag_3_746', 'diag_3_747', 'diag_3_750', 'diag_3_754', 'diag_3_755', 'diag_3_758', 'diag_3_759', 'diag_3_78', 'diag_3_801', 'diag_3_810', 'diag_3_814', 'diag_3_822', 'diag_3_826', 'diag_3_831', 'diag_3_834', 'diag_3_844', 'diag_3_845', 'diag_3_847', 'diag_3_851', 'diag_3_853', 'diag_3_854', 'diag_3_862', 'diag_3_864', 'diag_3_866', 'diag_3_868', 'diag_3_871', 'diag_3_872', 'diag_3_876', 'diag_3_877', 'diag_3_881', 'diag_3_883', 'diag_3_890', 'diag_3_906', 'diag_3_910', 'diag_3_911', 'diag_3_912', 'diag_3_918', 'diag_3_919', 'diag_3_921', 'diag_3_930', 'diag_3_933', 'diag_3_943', 'diag_3_945', 'diag_3_951', 'diag_3_952', 'diag_3_953', 'diag_3_955', 'diag_3_956', 'diag_3_962', 'diag_3_965', 'diag_3_966', 'diag_3_967', 'diag_3_969', 'diag_3_970', 'diag_3_972', 'diag_3_987', 'diag_3_989', 'diag_3_E815', 'diag_3_E817', 'diag_3_E818', 'diag_3_E819', 'diag_3_E822', 'diag_3_E825', 'diag_3_E826', 'diag_3_E828', 'diag_3_E850', 'diag_3_E852', 'diag_3_E853', 'diag_3_E858', 'diag_3_E861', 'diag_3_E882', 'diag_3_E883', 'diag_3_E887', 'diag_3_E905', 'diag_3_E906', 'diag_3_E916', 'diag_3_E917', 'diag_3_E924', 'diag_3_E930', 'diag_3_E941', 'diag_3_E943', 'diag_3_E949', 'diag_3_E966', 'diag_3_E980', 'diag_3_V02', 'diag_3_V03', 'diag_3_V13', 'diag_3_V16', 'diag_3_V23', 'diag_3_V57', 'diag_3_V61', 'diag_3_V66', 'diag_3_V86']
(_dystack pid=32908) 		These features carry no predictive signal and should be manually investigated.
(_dystack pid=32908) 		This is typically a feature which has the same value for all rows.
(_dystack pid=32908) 		These features do not need to be present at inference time.
(_dystack pid=32908) 	Unused Original Features (Count: 610): ['payer_code_MP', 'medical_specialty_ObstericsGynecologyGynecologicOnco', 'medical_specialty_Obstetrics', 'medical_specialty_Osteopath', 'medical_specialty_Podiatry', 'medical_specialty_Rheumatology', 'medical_specialty_SurgeryPlastic', 'medical_specialty_SurgicalSpecialty', 'diag_1_11', 'diag_1_115', 'diag_1_117', 'diag_1_133', 'diag_1_135', 'diag_1_136', 'diag_1_142', 'diag_1_155', 'diag_1_161', 'diag_1_163', 'diag_1_164', 'diag_1_171', 'diag_1_172', 'diag_1_179', 'diag_1_180', 'diag_1_184', 'diag_1_199', 'diag_1_203', 'diag_1_205', 'diag_1_214', 'diag_1_215', 'diag_1_229', 'diag_1_230', 'diag_1_233', 'diag_1_237', 'diag_1_239', 'diag_1_240', 'diag_1_244', 'diag_1_245', 'diag_1_2503', 'diag_1_25031', 'diag_1_25033', 'diag_1_25043', 'diag_1_25051', 'diag_1_25093', 'diag_1_251', 'diag_1_252', 'diag_1_253', 'diag_1_255', 'diag_1_263', 'diag_1_273', 'diag_1_277', 'diag_1_282', 'diag_1_283', 'diag_1_284', 'diag_1_288', 'diag_1_293', 'diag_1_299', 'diag_1_3', 'diag_1_305', 'diag_1_306', 'diag_1_307', 'diag_1_31', 'diag_1_314', 'diag_1_323', 'diag_1_325', 'diag_1_338', 'diag_1_34', 'diag_1_347', 'diag_1_350', 'diag_1_352', 'diag_1_354', 'diag_1_355', 'diag_1_356', 'diag_1_357', 'diag_1_360', 'diag_1_361', 'diag_1_368', 'diag_1_370', 'diag_1_376', 'diag_1_380', 'diag_1_384', 'diag_1_391', 'diag_1_394', 'diag_1_395', 'diag_1_405', 'diag_1_41', 'diag_1_416', 'diag_1_417', 'diag_1_423', 'diag_1_429', 'diag_1_430', 'diag_1_442', 'diag_1_452', 'diag_1_456', 'diag_1_462', 'diag_1_463', 'diag_1_464', 'diag_1_474', 'diag_1_475', 'diag_1_480', 'diag_1_483', 'diag_1_487', 'diag_1_492', 'diag_1_5', 'diag_1_508', 'diag_1_510', 'diag_1_512', 'diag_1_513', 'diag_1_514', 'diag_1_522', 'diag_1_54', 'diag_1_541', 'diag_1_567', 'diag_1_570', 'diag_1_579', 'diag_1_581', 'diag_1_586', 'diag_1_591', 'diag_1_594', 'diag_1_602', 'diag_1_608', 'diag_1_611', 'diag_1_614', 'diag_1_617', 'diag_1_619', 'diag_1_620', 'diag_1_622', 'diag_1_623', 'diag_1_633', 'diag_1_641', 'diag_1_647', 'diag_1_652', 'diag_1_655', 'diag_1_656', 'diag_1_658', 'diag_1_660', 'diag_1_664', 'diag_1_674', 'diag_1_686', 'diag_1_705', 'diag_1_708', 'diag_1_710', 'diag_1_717', 'diag_1_718', 'diag_1_725', 'diag_1_736', 'diag_1_751', 'diag_1_753', 'diag_1_759', 'diag_1_78', 'diag_1_797', 'diag_1_801', 'diag_1_802', 'diag_1_814', 'diag_1_815', 'diag_1_816', 'diag_1_82', 'diag_1_826', 'diag_1_833', 'diag_1_834', 'diag_1_847', 'diag_1_853', 'diag_1_861', 'diag_1_865', 'diag_1_866', 'diag_1_867', 'diag_1_873', 'diag_1_88', 'diag_1_880', 'diag_1_883', 'diag_1_915', 'diag_1_920', 'diag_1_933', 'diag_1_935', 'diag_1_939', 'diag_1_94', 'diag_1_942', 'diag_1_945', 'diag_1_952', 'diag_1_958', 'diag_1_959', 'diag_1_964', 'diag_1_965', 'diag_1_967', 'diag_1_970', 'diag_1_991', 'diag_1_992', 'diag_1_999', 'diag_1_V26', 'diag_1_V60', 'diag_1_V63', 'diag_1_V71', 'diag_2_110', 'diag_2_117', 'diag_2_136', 'diag_2_138', 'diag_2_140', 'diag_2_156', 'diag_2_172', 'diag_2_174', 'diag_2_179', 'diag_2_183', 'diag_2_185', 'diag_2_186', 'diag_2_191', 'diag_2_193', 'diag_2_200', 'diag_2_201', 'diag_2_203', 'diag_2_208', 'diag_2_215', 'diag_2_218', 'diag_2_220', 'diag_2_226', 'diag_2_233', 'diag_2_238', 'diag_2_241', 'diag_2_242', 'diag_2_246', 'diag_2_25011', 'diag_2_25022', 'diag_2_25033', 'diag_2_25053', 'diag_2_2509', 'diag_2_25091', 'diag_2_25093', 'diag_2_251', 'diag_2_253', 'diag_2_258', 'diag_2_260', 'diag_2_261', 'diag_2_262', 'diag_2_273', 'diag_2_274', 'diag_2_275', 'diag_2_277', 'diag_2_279', 'diag_2_282', 'diag_2_283', 'diag_2_298', 'diag_2_299', 'diag_2_301', 'diag_2_302', 'diag_2_307', 'diag_2_311', 'diag_2_314', 'diag_2_320', 'diag_2_322', 'diag_2_323', 'diag_2_327', 'diag_2_335', 'diag_2_336', 'diag_2_343', 'diag_2_345', 'diag_2_356', 'diag_2_358', 'diag_2_362', 'diag_2_365', 'diag_2_372', 'diag_2_377', 'diag_2_378', 'diag_2_380', 'diag_2_381', 'diag_2_398', 'diag_2_405', 'diag_2_421', 'diag_2_429', 'diag_2_431', 'diag_2_432', 'diag_2_446', 'diag_2_447', 'diag_2_451', 'diag_2_454', 'diag_2_455', 'diag_2_456', 'diag_2_457', 'diag_2_46', 'diag_2_460', 'diag_2_461', 'diag_2_462', 'diag_2_472', 'diag_2_474', 'diag_2_477', 'diag_2_478', 'diag_2_480', 'diag_2_485', 'diag_2_490', 'diag_2_494', 'diag_2_495', 'diag_2_500', 'diag_2_510', 'diag_2_512', 'diag_2_516', 'diag_2_517', 'diag_2_519', 'diag_2_524', 'diag_2_528', 'diag_2_53', 'diag_2_533', 'diag_2_54', 'diag_2_542', 'diag_2_543', 'diag_2_565', 'diag_2_566', 'diag_2_572', 'diag_2_573', 'diag_2_583', 'diag_2_586', 'diag_2_594', 'diag_2_596', 'diag_2_601', 'diag_2_602', 'diag_2_607', 'diag_2_608', 'diag_2_620', 'diag_2_621', 'diag_2_626', 'diag_2_646', 'diag_2_647', 'diag_2_649', 'diag_2_654', 'diag_2_656', 'diag_2_670', 'diag_2_691', 'diag_2_692', 'diag_2_693', 'diag_2_698', 'diag_2_712', 'diag_2_714', 'diag_2_716', 'diag_2_718', 'diag_2_719', 'diag_2_737', 'diag_2_752', 'diag_2_755', 'diag_2_783', 'diag_2_79', 'diag_2_791', 'diag_2_792', 'diag_2_802', 'diag_2_808', 'diag_2_810', 'diag_2_812', 'diag_2_813', 'diag_2_816', 'diag_2_821', 'diag_2_822', 'diag_2_825', 'diag_2_833', 'diag_2_836', 'diag_2_840', 'diag_2_842', 'diag_2_844', 'diag_2_851', 'diag_2_864', 'diag_2_865', 'diag_2_866', 'diag_2_869', 'diag_2_871', 'diag_2_88', 'diag_2_883', 'diag_2_892', 'diag_2_907', 'diag_2_909', 'diag_2_911', 'diag_2_916', 'diag_2_919', 'diag_2_921', 'diag_2_923', 'diag_2_934', 'diag_2_948', 'diag_2_955', 'diag_2_958', 'diag_2_972', 'diag_2_980', 'diag_2_989', 'diag_2_999', 'diag_2_E826', 'diag_2_E870', 'diag_2_E879', 'diag_2_E881', 'diag_2_E884', 'diag_2_E917', 'diag_2_E918', 'diag_2_E919', 'diag_2_E927', 'diag_2_E933', 'diag_2_E934', 'diag_2_E935', 'diag_2_E942', 'diag_2_V02', 'diag_2_V08', 'diag_2_V14', 'diag_2_V44', 'diag_2_V49', 'diag_2_V63', 'diag_2_V65', 'diag_3_155', 'diag_3_156', 'diag_3_157', 'diag_3_158', 'diag_3_173', 'diag_3_179', 'diag_3_180', 'diag_3_183', 'diag_3_188', 'diag_3_191', 'diag_3_192', 'diag_3_199', 'diag_3_201', 'diag_3_202', 'diag_3_208', 'diag_3_220', 'diag_3_225', 'diag_3_228', 'diag_3_241', 'diag_3_25013', 'diag_3_2502', 'diag_3_25043', 'diag_3_2507', 'diag_3_25083', 'diag_3_2509', 'diag_3_25091', 'diag_3_25093', 'diag_3_251', 'diag_3_252', 'diag_3_256', 'diag_3_262', 'diag_3_277', 'diag_3_279', 'diag_3_281', 'diag_3_283', 'diag_3_289', 'diag_3_297', 'diag_3_298', 'diag_3_306', 'diag_3_313', 'diag_3_317', 'diag_3_332', 'diag_3_334', 'diag_3_335', 'diag_3_336', 'diag_3_343', 'diag_3_347', 'diag_3_353', 'diag_3_354', 'diag_3_355', 'diag_3_356', 'diag_3_365', 'diag_3_368', 'diag_3_369', 'diag_3_372', 'diag_3_377', 'diag_3_384', 'diag_3_389', 'diag_3_391', 'diag_3_394', 'diag_3_420', 'diag_3_423', 'diag_3_431', 'diag_3_442', 'diag_3_444', 'diag_3_446', 'diag_3_451', 'diag_3_457', 'diag_3_465', 'diag_3_477', 'diag_3_494', 'diag_3_495', 'diag_3_510', 'diag_3_519', 'diag_3_528', 'diag_3_529', 'diag_3_53', 'diag_3_531', 'diag_3_532', 'diag_3_533', 'diag_3_537', 'diag_3_54', 'diag_3_550', 'diag_3_555', 'diag_3_568', 'diag_3_57', 'diag_3_572', 'diag_3_573', 'diag_3_575', 'diag_3_576', 'diag_3_582', 'diag_3_586', 'diag_3_592', 'diag_3_594', 'diag_3_604', 'diag_3_605', 'diag_3_607', 'diag_3_611', 'diag_3_619', 'diag_3_620', 'diag_3_621', 'diag_3_623', 'diag_3_626', 'diag_3_627', 'diag_3_644', 'diag_3_646', 'diag_3_649', 'diag_3_653', 'diag_3_654', 'diag_3_656', 'diag_3_660', 'diag_3_661', 'diag_3_663', 'diag_3_664', 'diag_3_670', 'diag_3_692', 'diag_3_693', 'diag_3_695', 'diag_3_696', 'diag_3_697', 'diag_3_701', 'diag_3_703', 'diag_3_708', 'diag_3_709', 'diag_3_717', 'diag_3_718', 'diag_3_720', 'diag_3_722', 'diag_3_725', 'diag_3_726', 'diag_3_745', 'diag_3_751', 'diag_3_752', 'diag_3_753', 'diag_3_756', 'diag_3_791', 'diag_3_792', 'diag_3_793', 'diag_3_796', 'diag_3_797', 'diag_3_800', 'diag_3_807', 'diag_3_808', 'diag_3_815', 'diag_3_816', 'diag_3_823', 'diag_3_836', 'diag_3_837', 'diag_3_838', 'diag_3_840', 'diag_3_841', 'diag_3_842', 'diag_3_850', 'diag_3_861', 'diag_3_865', 'diag_3_867', 'diag_3_875', 'diag_3_880', 'diag_3_882', 'diag_3_891', 'diag_3_892', 'diag_3_907', 'diag_3_916', 'diag_3_917', 'diag_3_922', 'diag_3_924', 'diag_3_934', 'diag_3_94', 'diag_3_948', 'diag_3_959', 'diag_3_971', 'diag_3_E812', 'diag_3_E816', 'diag_3_E870', 'diag_3_E884', 'diag_3_E892', 'diag_3_E894', 'diag_3_E900', 'diag_3_E912', 'diag_3_E919', 'diag_3_E920', 'diag_3_E922', 'diag_3_E927', 'diag_3_E928', 'diag_3_E929', 'diag_3_E931', 'diag_3_E933', 'diag_3_E934', 'diag_3_E935', 'diag_3_E938', 'diag_3_E944', 'diag_3_E946', 'diag_3_E950', 'diag_3_E956', 'diag_3_V06', 'diag_3_V09', 'diag_3_V11', 'diag_3_V44', 'diag_3_V46', 'diag_3_V53', 'diag_3_V55', 'diag_3_V62', 'diag_3_V63', 'diag_3_V65', 'diag_3_V70', 'nateglinide_Up', 'chlorpropamide_Steady', 'tolbutamide_Steady', 'acarbose_Up', 'miglitol_Steady', 'tolazamide_Steady', 'group_diag_1_na', 'group_diag_2_na', 'group_diag_3_na']
(_dystack pid=32908) 		These features were not used to generate any of the output features. Add a feature generator compatible with these features to utilize them.
(_dystack pid=32908) 		Features can also be unused if they carry very little information, such as being categorical but having almost entirely unique values or being duplicates of other features.
(_dystack pid=32908) 		These features do not need to be present at inference time.
(_dystack pid=32908) 		('float', []) : 610 | ['payer_code_MP', 'medical_specialty_ObstericsGynecologyGynecologicOnco', 'medical_specialty_Obstetrics', 'medical_specialty_Osteopath', 'medical_specialty_Podiatry', ...]
(_dystack pid=32908) 	Types of features in original data (raw dtype, special dtypes):
(_dystack pid=32908) 		('float', []) : 1065 | ['time_in_hospital', 'num_lab_procedures', 'num_procedures', 'num_medications', 'number_outpatient', ...]
(_dystack pid=32908) 	Types of features in processed data (raw dtype, special dtypes):
(_dystack pid=32908) 		('float', [])     :    8 | ['time_in_hospital', 'num_lab_procedures', 'num_procedures', 'num_medications', 'number_outpatient', ...]
(_dystack pid=32908) 		('int', ['bool']) : 1057 | ['race_AfricanAmerican', 'race_Asian', 'race_Caucasian', 'race_Hispanic', 'race_Other', ...]
(_dystack pid=32908) 	11.9s = Fit runtime
(_dystack pid=32908) 	1065 features in original data used to generate 1065 features in processed data.
(_dystack pid=32908) 	Train Data (Processed) Memory Usage: 12.52 MB (0.5% of available memory)
(_dystack pid=32908) Data preprocessing and feature engineering runtime = 12.18s ...
(_dystack pid=32908) AutoGluon will gauge predictive performance using evaluation metric: 'roc_auc'
(_dystack pid=32908) 	This metric expects predicted probabilities rather than predicted class labels, so you'll need to use predict_proba() instead of predict()
(_dystack pid=32908) 	To change this, specify the eval_metric parameter of Predictor()
(_dystack pid=32908) Large model count detected (112 configs) ... Only displaying the first 3 models of each family. To see all, set `verbosity=3`.
(_dystack pid=32908) User-specified model hyperparameters to be fit:
(_dystack pid=32908) {
(_dystack pid=32908) 	'NN_TORCH': [{}, {'activation': 'elu', 'dropout_prob': 0.10077639529843717, 'hidden_size': 108, 'learning_rate': 0.002735937344002146, 'num_layers': 4, 'use_batchnorm': True, 'weight_decay': 1.356433327634438e-12, 'ag_args': {'name_suffix': '_r79', 'priority': -2}}, {'activation': 'elu', 'dropout_prob': 0.11897478034205347, 'hidden_size': 213, 'learning_rate': 0.0010474382260641949, 'num_layers': 4, 'use_batchnorm': False, 'weight_decay': 5.594471067786272e-10, 'ag_args': {'name_suffix': '_r22', 'priority': -7}}],
(_dystack pid=32908) 	'GBM': [{'extra_trees': True, 'ag_args': {'name_suffix': 'XT'}}, {}, {'learning_rate': 0.03, 'num_leaves': 128, 'feature_fraction': 0.9, 'min_data_in_leaf': 3, 'ag_args': {'name_suffix': 'Large', 'priority': 0, 'hyperparameter_tune_kwargs': None}}],
(_dystack pid=32908) 	'CAT': [{}, {'depth': 6, 'grow_policy': 'SymmetricTree', 'l2_leaf_reg': 2.1542798306067823, 'learning_rate': 0.06864209415792857, 'max_ctr_complexity': 4, 'one_hot_max_size': 10, 'ag_args': {'name_suffix': '_r177', 'priority': -1}}, {'depth': 8, 'grow_policy': 'Depthwise', 'l2_leaf_reg': 2.7997999596449104, 'learning_rate': 0.031375015734637225, 'max_ctr_complexity': 2, 'one_hot_max_size': 3, 'ag_args': {'name_suffix': '_r9', 'priority': -5}}],
(_dystack pid=32908) 	'XGB': [{}, {'colsample_bytree': 0.6917311125174739, 'enable_categorical': False, 'learning_rate': 0.018063876087523967, 'max_depth': 10, 'min_child_weight': 0.6028633586934382, 'ag_args': {'name_suffix': '_r33', 'priority': -8}}, {'colsample_bytree': 0.6628423832084077, 'enable_categorical': False, 'learning_rate': 0.08775715546881824, 'max_depth': 5, 'min_child_weight': 0.6294123374222513, 'ag_args': {'name_suffix': '_r89', 'priority': -16}}],
(_dystack pid=32908) 	'FASTAI': [{}, {'bs': 256, 'emb_drop': 0.5411770367537934, 'epochs': 43, 'layers': [800, 400], 'lr': 0.01519848858318159, 'ps': 0.23782946566604385, 'ag_args': {'name_suffix': '_r191', 'priority': -4}}, {'bs': 2048, 'emb_drop': 0.05070411322605811, 'epochs': 29, 'layers': [200, 100], 'lr': 0.08974235041576624, 'ps': 0.10393466140748028, 'ag_args': {'name_suffix': '_r102', 'priority': -11}}],
(_dystack pid=32908) 	'RF': [{'criterion': 'gini', 'ag_args': {'name_suffix': 'Gini', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'entropy', 'ag_args': {'name_suffix': 'Entr', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'squared_error', 'ag_args': {'name_suffix': 'MSE', 'problem_types': ['regression', 'quantile']}}],
(_dystack pid=32908) 	'XT': [{'criterion': 'gini', 'ag_args': {'name_suffix': 'Gini', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'entropy', 'ag_args': {'name_suffix': 'Entr', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'squared_error', 'ag_args': {'name_suffix': 'MSE', 'problem_types': ['regression', 'quantile']}}],
(_dystack pid=32908) 	'KNN': [{'weights': 'uniform', 'ag_args': {'name_suffix': 'Unif'}}, {'weights': 'distance', 'ag_args': {'name_suffix': 'Dist'}}],
(_dystack pid=32908) }
(_dystack pid=32908) AutoGluon will fit 3 stack levels (L1 to L3) ...
(_dystack pid=32908) Fitting 110 L1 models, fit_strategy="sequential" ...
(_dystack pid=32908) Fitting model: KNeighborsUnif_BAG_L1 ... Training model for up to 390.73s of the 879.36s of remaining time.
(_dystack pid=32908) 	0.5801	 = Validation score   (roc_auc)
(_dystack pid=32908) 	0.08s	 = Training   runtime
(_dystack pid=32908) 	0.55s	 = Validation runtime
(_dystack pid=32908) Fitting model: KNeighborsDist_BAG_L1 ... Training model for up to 386.38s of the 875.01s of remaining time.
(_dystack pid=32908) 	0.58	 = Validation score   (roc_auc)
(_dystack pid=32908) 	0.08s	 = Training   runtime
(_dystack pid=32908) 	0.61s	 = Validation runtime
(_dystack pid=32908) Fitting model: LightGBMXT_BAG_L1 ... Training model for up to 385.50s of the 874.13s of remaining time.
(_dystack pid=32908) 	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 13.24% memory usage per fold, 52.94%/80.00% total).
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=13.24%)
(_dystack pid=32908) 	0.6936	 = Validation score   (roc_auc)
(_dystack pid=32908) 	10.9s	 = Training   runtime
(_dystack pid=32908) 	0.33s	 = Validation runtime
(_dystack pid=32908) Fitting model: LightGBM_BAG_L1 ... Training model for up to 369.92s of the 858.54s of remaining time.
(_dystack pid=32908) 	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 11.69% memory usage per fold, 46.77%/80.00% total).
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=11.69%)
(_dystack pid=32908) 	0.6921	 = Validation score   (roc_auc)
(_dystack pid=32908) 	18.42s	 = Training   runtime
(_dystack pid=32908) 	0.52s	 = Validation runtime
(_dystack pid=32908) Fitting model: RandomForestGini_BAG_L1 ... Training model for up to 346.88s of the 835.50s of remaining time.
(_dystack pid=32908) 	0.6826	 = Validation score   (roc_auc)
(_dystack pid=32908) 	10.74s	 = Training   runtime
(_dystack pid=32908) 	6.92s	 = Validation runtime
(_dystack pid=32908) Fitting model: RandomForestEntr_BAG_L1 ... Training model for up to 327.93s of the 816.56s of remaining time.
(_dystack pid=32908) 	0.6837	 = Validation score   (roc_auc)
(_dystack pid=32908) 	9.97s	 = Training   runtime
(_dystack pid=32908) 	6.6s	 = Validation runtime
(_dystack pid=32908) Fitting model: CatBoost_BAG_L1 ... Training model for up to 310.30s of the 798.92s of remaining time.
(_dystack pid=32908) 	Warning: Potentially not enough memory to safely train model. Estimated to require 1.483 GB out of 1.744 GB available memory (85.036%)... (100.000% of avail memory is the max safe size)
(_dystack pid=32908) 	To avoid this warning, specify the model hyperparameter "ag.max_memory_usage_ratio" to a larger value (currently 1.0, set to >=1.18 to avoid the warning)
(_dystack pid=32908) 		To set the same value for all models, do the following when calling predictor.fit: `predictor.fit(..., ag_args_fit={"ag.max_memory_usage_ratio": VALUE})`
(_dystack pid=32908) 		Setting "ag.max_memory_usage_ratio" to values above 1 may result in out-of-memory errors. You may consider using a machine with more memory as a safer alternative.
(_dystack pid=32908) 	Memory not enough to fit 8 folds in parallel. Will train 1 folds in parallel instead (Estimated 85.04% memory usage per fold, 85.04%/80.00% total).
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (1 workers, per: cpus=6, gpus=0, memory=85.04%)
(_dystack pid=32908) 		Switching to pseudo sequential ParallelFoldFittingStrategy to avoid Python memory leakage.
(_dystack pid=32908) 		Overrule this behavior by setting fold_fitting_strategy to 'sequential_local' in ag_args_ensemble when when calling `predictor.fit`
(_ray_fit pid=16808) 	Warning: Potentially not enough memory to safely train model. Estimated to require 1.461 GB out of 1.515 GB available memory (96.416%)... (100.000% of avail memory is the max safe size)
(_ray_fit pid=16808) 	To avoid this warning, specify the model hyperparameter "ag.max_memory_usage_ratio" to a larger value (currently 1.0, set to >=1.34 to avoid the warning)
(_ray_fit pid=16808) 		To set the same value for all models, do the following when calling predictor.fit: `predictor.fit(..., ag_args_fit={"ag.max_memory_usage_ratio": VALUE})`
(_ray_fit pid=16808) 		Setting "ag.max_memory_usage_ratio" to values above 1 may result in out-of-memory errors. You may consider using a machine with more memory as a safer alternative.
(_ray_fit pid=1152) 	Warning: Potentially not enough memory to safely train model. Estimated to require 1.461 GB out of 1.901 GB available memory (76.844%)... (100.000% of avail memory is the max safe size)
(_ray_fit pid=1152) 	To avoid this warning, specify the model hyperparameter "ag.max_memory_usage_ratio" to a larger value (currently 1.0, set to >=1.07 to avoid the warning)
(_ray_fit pid=1152) 		To set the same value for all models, do the following when calling predictor.fit: `predictor.fit(..., ag_args_fit={"ag.max_memory_usage_ratio": VALUE})`
(_ray_fit pid=1152) 		Setting "ag.max_memory_usage_ratio" to values above 1 may result in out-of-memory errors. You may consider using a machine with more memory as a safer alternative.
(_ray_fit pid=22484) 	Warning: Potentially not enough memory to safely train model. Estimated to require 1.461 GB out of 1.702 GB available memory (85.855%)... (100.000% of avail memory is the max safe size)
(_ray_fit pid=22484) 	To avoid this warning, specify the model hyperparameter "ag.max_memory_usage_ratio" to a larger value (currently 1.0, set to >=1.19 to avoid the warning)
(_ray_fit pid=22484) 		To set the same value for all models, do the following when calling predictor.fit: `predictor.fit(..., ag_args_fit={"ag.max_memory_usage_ratio": VALUE})`
(_ray_fit pid=22484) 		Setting "ag.max_memory_usage_ratio" to values above 1 may result in out-of-memory errors. You may consider using a machine with more memory as a safer alternative.
(_ray_fit pid=27396) 	Warning: Potentially not enough memory to safely train model. Estimated to require 1.461 GB out of 1.553 GB available memory (94.088%)... (100.000% of avail memory is the max safe size)
(_ray_fit pid=27396) 	To avoid this warning, specify the model hyperparameter "ag.max_memory_usage_ratio" to a larger value (currently 1.0, set to >=1.30 to avoid the warning)
(_ray_fit pid=27396) 		To set the same value for all models, do the following when calling predictor.fit: `predictor.fit(..., ag_args_fit={"ag.max_memory_usage_ratio": VALUE})`
(_ray_fit pid=27396) 		Setting "ag.max_memory_usage_ratio" to values above 1 may result in out-of-memory errors. You may consider using a machine with more memory as a safer alternative.
(_ray_fit pid=27396) Iteration 6: Low available memory (<512 MB). Available memory: 489.45 MB, Estimated model size: 12.73 MB.
(_ray_fit pid=27396) Early stopping model prior to optimal result to avoid OOM error. Please increase available memory to avoid subpar model quality.
(_ray_fit pid=27396) 	Ran low on memory, early stopping on iteration 6.
(_dystack pid=32908) 	0.6814	 = Validation score   (roc_auc)
(_dystack pid=32908) 	138.48s	 = Training   runtime
(_dystack pid=32908) 	1.48s	 = Validation runtime
(_dystack pid=32908) Fitting model: ExtraTreesGini_BAG_L1 ... Training model for up to 165.63s of the 654.25s of remaining time.
(_dystack pid=32908) 	0.6834	 = Validation score   (roc_auc)
(_dystack pid=32908) 	11.24s	 = Training   runtime
(_dystack pid=32908) 	6.37s	 = Validation runtime
(_dystack pid=32908) Fitting model: ExtraTreesEntr_BAG_L1 ... Training model for up to 146.99s of the 635.61s of remaining time.
(_dystack pid=32908) 	0.6774	 = Validation score   (roc_auc)
(_dystack pid=32908) 	11.26s	 = Training   runtime
(_dystack pid=32908) 	7.4s	 = Validation runtime
(_dystack pid=32908) Fitting model: NeuralNetFastAI_BAG_L1 ... Training model for up to 127.45s of the 616.07s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=4.25%)
(_ray_fit pid=26056) No improvement since epoch 0: early stopping
(_dystack pid=32908) 	0.6141	 = Validation score   (roc_auc)
(_dystack pid=32908) 	68.74s	 = Training   runtime
(_dystack pid=32908) 	1.94s	 = Validation runtime
(_dystack pid=32908) Fitting model: XGBoost_BAG_L1 ... Training model for up to 50.68s of the 539.30s of remaining time.
(_dystack pid=32908) 	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 10.06% memory usage per fold, 40.24%/80.00% total).
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=10.06%)
(_dystack pid=32908) 	0.6877	 = Validation score   (roc_auc)
(_dystack pid=32908) 	46.89s	 = Training   runtime
(_dystack pid=32908) 	0.74s	 = Validation runtime
(_ray_fit pid=27516) No improvement since epoch 0: early stopping [repeated 7x across cluster] (Ray deduplicates logs by default. Set RAY_DEDUP_LOGS=0 to disable log deduplication, or see https://docs.ray.io/en/master/ray-observability/user-guides/configure-logging.html#log-deduplication for more options.)
(_dystack pid=32908) Fitting model: WeightedEnsemble_L2 ... Training model for up to 360.00s of the 484.12s of remaining time.
(_dystack pid=32908) 	Ensemble Weights: {'ExtraTreesGini_BAG_L1': 0.25, 'LightGBMXT_BAG_L1': 0.208, 'LightGBM_BAG_L1': 0.167, 'CatBoost_BAG_L1': 0.167, 'RandomForestGini_BAG_L1': 0.083, 'KNeighborsDist_BAG_L1': 0.042, 'RandomForestEntr_BAG_L1': 0.042, 'XGBoost_BAG_L1': 0.042}
(_dystack pid=32908) 	0.7013	 = Validation score   (roc_auc)
(_dystack pid=32908) 	0.89s	 = Training   runtime
(_dystack pid=32908) 	0.0s	 = Validation runtime
(_dystack pid=32908) Fitting 108 L2 models, fit_strategy="sequential" ...
(_dystack pid=32908) Fitting model: LightGBMXT_BAG_L2 ... Training model for up to 322.05s of the 482.86s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=8.61%)
(_dystack pid=32908) 	0.6942	 = Validation score   (roc_auc)
(_dystack pid=32908) 	6.3s	 = Training   runtime
(_dystack pid=32908) 	0.54s	 = Validation runtime
(_dystack pid=32908) Fitting model: LightGBM_BAG_L2 ... Training model for up to 305.69s of the 466.50s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=6.34%)
(_dystack pid=32908) 	0.6928	 = Validation score   (roc_auc)
(_dystack pid=32908) 	5.21s	 = Training   runtime
(_dystack pid=32908) 	0.48s	 = Validation runtime
(_dystack pid=32908) Fitting model: RandomForestGini_BAG_L2 ... Training model for up to 291.12s of the 451.93s of remaining time.
(_dystack pid=32908) 	0.6872	 = Validation score   (roc_auc)
(_dystack pid=32908) 	6.57s	 = Training   runtime
(_dystack pid=32908) 	6.35s	 = Validation runtime
(_dystack pid=32908) Fitting model: RandomForestEntr_BAG_L2 ... Training model for up to 277.06s of the 437.87s of remaining time.
(_dystack pid=32908) 	0.6939	 = Validation score   (roc_auc)
(_dystack pid=32908) 	7.45s	 = Training   runtime
(_dystack pid=32908) 	10.16s	 = Validation runtime
(_dystack pid=32908) Fitting model: CatBoost_BAG_L2 ... Training model for up to 258.50s of the 419.31s of remaining time.
(_dystack pid=32908) 	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 32.49% memory usage per fold, 64.97%/80.00% total).
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=32.49%)
(_ray_fit pid=32476) Iteration 290: Low available memory (<512 MB). Available memory: 465.66 MB, Estimated model size: 17.73 MB.
(_ray_fit pid=32476) Early stopping model prior to optimal result to avoid OOM error. Please increase available memory to avoid subpar model quality.
(_ray_fit pid=32476) 	Ran low on memory, early stopping on iteration 290.
(_dystack pid=32908) 	0.6982	 = Validation score   (roc_auc)
(_dystack pid=32908) 	56.72s	 = Training   runtime
(_dystack pid=32908) 	1.61s	 = Validation runtime
(_dystack pid=32908) Fitting model: ExtraTreesGini_BAG_L2 ... Training model for up to 194.85s of the 355.66s of remaining time.
(_ray_fit pid=31048) Iteration 280: Low available memory (<512 MB). Available memory: 353.31 MB, Estimated model size: 18.46 MB.
(_ray_fit pid=31048) Early stopping model prior to optimal result to avoid OOM error. Please increase available memory to avoid subpar model quality.
(_ray_fit pid=31048) 	Ran low on memory, early stopping on iteration 280.
(_dystack pid=32908) 	0.6898	 = Validation score   (roc_auc)
(_dystack pid=32908) 	9.48s	 = Training   runtime
(_dystack pid=32908) 	6.47s	 = Validation runtime
(_dystack pid=32908) Fitting model: ExtraTreesEntr_BAG_L2 ... Training model for up to 178.05s of the 338.87s of remaining time.
(_dystack pid=32908) 	0.6918	 = Validation score   (roc_auc)
(_dystack pid=32908) 	8.9s	 = Training   runtime
(_dystack pid=32908) 	5.95s	 = Validation runtime
(_dystack pid=32908) Fitting model: NeuralNetFastAI_BAG_L2 ... Training model for up to 162.42s of the 323.23s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.07%)
(_ray_fit pid=952) No improvement since epoch 0: early stopping
(_dystack pid=32908) 	0.6209	 = Validation score   (roc_auc)
(_dystack pid=32908) 	69.11s	 = Training   runtime
(_dystack pid=32908) 	1.78s	 = Validation runtime
(_dystack pid=32908) Fitting model: XGBoost_BAG_L2 ... Training model for up to 82.89s of the 243.70s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=8.99%)
(_dystack pid=32908) 	0.6895	 = Validation score   (roc_auc)
(_dystack pid=32908) 	38.12s	 = Training   runtime
(_dystack pid=32908) 	0.59s	 = Validation runtime
(_dystack pid=32908) Fitting model: NeuralNetTorch_BAG_L2 ... Training model for up to 30.20s of the 191.01s of remaining time.
(_ray_fit pid=35136) No improvement since epoch 0: early stopping [repeated 7x across cluster]
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.29%)
(_ray_fit pid=17260) 	Ran out of time, stopping training early. (Stopping on epoch 11)
(_dystack pid=32908) 	0.6951	 = Validation score   (roc_auc)
(_dystack pid=32908) 	30.91s	 = Training   runtime
(_dystack pid=32908) 	1.08s	 = Validation runtime
(_dystack pid=32908) Fitting model: WeightedEnsemble_L3 ... Training model for up to 360.00s of the 150.06s of remaining time.
(_dystack pid=32908) 	Ensemble Weights: {'CatBoost_BAG_L2': 0.5, 'RandomForestEntr_BAG_L2': 0.167, 'ExtraTreesEntr_BAG_L2': 0.167, 'NeuralNetTorch_BAG_L2': 0.125, 'ExtraTreesGini_BAG_L2': 0.042}
(_dystack pid=32908) 	0.6994	 = Validation score   (roc_auc)
(_dystack pid=32908) 	1.07s	 = Training   runtime
(_dystack pid=32908) 	0.0s	 = Validation runtime
(_dystack pid=32908) Fitting 108 L3 models, fit_strategy="sequential" ...
(_dystack pid=32908) Fitting model: LightGBMXT_BAG_L3 ... Training model for up to 148.93s of the 148.39s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=5.04%)
(_ray_fit pid=13300) 	Ran out of time, stopping training early. (Stopping on epoch 12) [repeated 6x across cluster]
(_dystack pid=32908) 	0.6932	 = Validation score   (roc_auc)
(_dystack pid=32908) 	4.76s	 = Training   runtime
(_dystack pid=32908) 	0.53s	 = Validation runtime
(_dystack pid=32908) Fitting model: LightGBM_BAG_L3 ... Training model for up to 135.02s of the 134.48s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=5.36%)
(_dystack pid=32908) 	0.6911	 = Validation score   (roc_auc)
(_dystack pid=32908) 	5.12s	 = Training   runtime
(_dystack pid=32908) 	0.51s	 = Validation runtime
(_dystack pid=32908) Fitting model: RandomForestGini_BAG_L3 ... Training model for up to 121.23s of the 120.69s of remaining time.
(_dystack pid=32908) 	0.687	 = Validation score   (roc_auc)
(_dystack pid=32908) 	6.62s	 = Training   runtime
(_dystack pid=32908) 	5.53s	 = Validation runtime
(_dystack pid=32908) Fitting model: RandomForestEntr_BAG_L3 ... Training model for up to 108.11s of the 107.57s of remaining time.
(_dystack pid=32908) 	0.6861	 = Validation score   (roc_auc)
(_dystack pid=32908) 	7.17s	 = Training   runtime
(_dystack pid=32908) 	5.43s	 = Validation runtime
(_dystack pid=32908) Fitting model: CatBoost_BAG_L3 ... Training model for up to 94.80s of the 94.26s of remaining time.
(_dystack pid=32908) 	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 28.45% memory usage per fold, 56.91%/80.00% total).
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=28.45%)
(_dystack pid=32908) 	0.6971	 = Validation score   (roc_auc)
(_dystack pid=32908) 	45.12s	 = Training   runtime
(_dystack pid=32908) 	1.35s	 = Validation runtime
(_dystack pid=32908) Fitting model: ExtraTreesGini_BAG_L3 ... Training model for up to 44.97s of the 44.43s of remaining time.
(_dystack pid=32908) 	0.6884	 = Validation score   (roc_auc)
(_dystack pid=32908) 	7.48s	 = Training   runtime
(_dystack pid=32908) 	5.59s	 = Validation runtime
(_dystack pid=32908) Fitting model: ExtraTreesEntr_BAG_L3 ... Training model for up to 31.20s of the 30.66s of remaining time.
(_dystack pid=32908) 	0.6883	 = Validation score   (roc_auc)
(_dystack pid=32908) 	8.26s	 = Training   runtime
(_dystack pid=32908) 	5.63s	 = Validation runtime
(_dystack pid=32908) Fitting model: NeuralNetFastAI_BAG_L3 ... Training model for up to 16.51s of the 15.97s of remaining time.
(_dystack pid=32908) 	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.51%)
(_dystack pid=32908) 	Time limit exceeded... Skipping NeuralNetFastAI_BAG_L3.
(_dystack pid=32908) Fitting model: WeightedEnsemble_L4 ... Training model for up to 360.00s of the -6.83s of remaining time.
(_dystack pid=32908) 	Ensemble Weights: {'LightGBMXT_BAG_L1': 0.28, 'CatBoost_BAG_L2': 0.16, 'CatBoost_BAG_L1': 0.12, 'RandomForestEntr_BAG_L2': 0.12, 'ExtraTreesEntr_BAG_L2': 0.12, 'ExtraTreesGini_BAG_L3': 0.08, 'KNeighborsDist_BAG_L1': 0.04, 'XGBoost_BAG_L1': 0.04, 'CatBoost_BAG_L3': 0.04}
(_dystack pid=32908) 	0.7005	 = Validation score   (roc_auc)
(_dystack pid=32908) 	1.19s	 = Training   runtime
(_dystack pid=32908) 	0.0s	 = Validation runtime
(_dystack pid=32908) AutoGluon training complete, total runtime = 899.64s ... Best model: WeightedEnsemble_L2 | Estimated inference throughput: 259.6 rows/s (1464 batch size)
(_dystack pid=32908) TabularPredictor saved. To load, use: predictor = TabularPredictor.load("c:\studium\DAV Ausbildung\CADS\Immersion\ads3_exam\03_scripts\autogluon_diabetes_auc\ds_sub_fit\sub_fit_ho")
(_dystack pid=32908) Unhandled error (suppress with 'RAY_IGNORE_UNHANDLED_ERRORS=1'): The worker died unexpectedly while executing this task. Check python-core-worker-*.log files for more information.
(_dystack pid=32908) Deleting DyStack predictor artifacts (clean_up_fits=True) ...
Leaderboard on holdout data (DyStack):
                      model  score_holdout  score_val eval_metric  pred_time_test  pred_time_val    fit_time  pred_time_test_marginal  pred_time_val_marginal  fit_time_marginal  stack_level  can_infer  fit_order
0           LightGBM_BAG_L1       0.702035   0.692062     roc_auc        0.720246       0.524789   18.419194                 0.720246                0.524789          18.419194            1       True          4
1            XGBoost_BAG_L1       0.700820   0.687669     roc_auc        1.123045       0.739026   46.888627                 1.123045                0.739026          46.888627            1       True         11
2           LightGBM_BAG_L2       0.700251   0.692762     roc_auc        9.176113      33.953432  332.016193                 0.492544                0.482172           5.213061            2       True         14
3            XGBoost_BAG_L2       0.699033   0.689484     roc_auc        9.603825      34.059568  364.922395                 0.920256                0.588307          38.119262            2       True         21
4         LightGBMXT_BAG_L1       0.698827   0.693619     roc_auc        0.971671       0.333140   10.899239                 0.971671                0.333140          10.899239            1       True          3
5         LightGBMXT_BAG_L2       0.698107   0.694226     roc_auc        9.206035      34.011583  333.101941                 0.522466                0.540322           6.298809            2       True         13
6           CatBoost_BAG_L2       0.697666   0.698169     roc_auc        9.365991      35.083744  383.523052                 0.682422                1.612483          56.719920            2       True         17
7       WeightedEnsemble_L4       0.697211   0.700509     roc_auc       16.088237      75.411415  619.356147                 0.034240                0.000000           1.187552            4       True         31
8       WeightedEnsemble_L2       0.696547   0.701327     roc_auc        6.046668      23.579976  247.610606                 0.025848                0.002013           0.887570            2       True         12
9           CatBoost_BAG_L3       0.695823   0.697128     roc_auc       15.596728      69.818422  610.692262                 0.853104                1.345728          45.115420            3       True         28
10        LightGBMXT_BAG_L3       0.695640   0.693224     roc_auc       15.478703      69.005402  570.340166                 0.735079                0.532707           4.763324            3       True         24
11      WeightedEnsemble_L3       0.695121   0.699397     roc_auc       11.411233      58.739377  441.333209                 0.030765                0.003989           1.065155            3       True         23
12    NeuralNetTorch_BAG_L2       0.694657   0.695138     roc_auc        9.619292      34.548492  357.708633                 0.935724                1.077231          30.905501            2       True         22
13          CatBoost_BAG_L1       0.694515   0.681403     roc_auc        1.753284       1.479760  138.480240                 1.753284                1.479760         138.480240            1       True          7
14          LightGBM_BAG_L3       0.694351   0.691087     roc_auc       15.433444      68.987542  570.697489                 0.689819                0.514848           5.120647            3       True         25
15    ExtraTreesEntr_BAG_L3       0.693597   0.688321     roc_auc       15.270735      74.106896  573.833061                 0.527111                5.634202           8.256219            3       True         30
16  RandomForestEntr_BAG_L2       0.692863   0.693938     roc_auc        9.030982      43.630247  334.254858                 0.347413               10.158986           7.451726            2       True         16
17  RandomForestEntr_BAG_L3       0.692686   0.686109     roc_auc       15.091510      73.904293  572.747903                 0.347885                5.431598           7.171061            3       True         27
18  RandomForestGini_BAG_L3       0.691675   0.687028     roc_auc       15.188117      74.001036  572.201475                 0.444493                5.528342           6.624633            3       True         26
19    ExtraTreesGini_BAG_L3       0.685558   0.688437     roc_auc       15.200894      74.065688  573.053175                 0.457269                5.592993           7.476333            3       True         29
20    ExtraTreesGini_BAG_L2       0.684813   0.689820     roc_auc        9.036015      39.937354  336.287470                 0.352446                6.466093           9.484338            2       True         18
21  RandomForestGini_BAG_L2       0.684480   0.687225     roc_auc        9.007171      39.819417  333.372889                 0.323603                6.348156           6.569757            2       True         15
22    ExtraTreesEntr_BAG_L2       0.683113   0.691798     roc_auc        9.062463      39.420594  335.706569                 0.378895                5.949334           8.903437            2       True         19
23  RandomForestGini_BAG_L1       0.676770   0.682589     roc_auc        0.447961       6.922239   10.741470                 0.447961                6.922239          10.741470            1       True          5
24  RandomForestEntr_BAG_L1       0.674646   0.683656     roc_auc        0.433249       6.597428    9.971981                 0.433249                6.597428           9.971981            1       True          6
25    ExtraTreesGini_BAG_L1       0.673604   0.683448     roc_auc        0.460975       6.374763   11.239830                 0.460975                6.374763          11.239830            1       True          8
26    ExtraTreesEntr_BAG_L1       0.666480   0.677444     roc_auc        0.429228       7.400769   11.257731                 0.429228                7.400769          11.257731            1       True          9
27   NeuralNetFastAI_BAG_L2       0.661369   0.620925     roc_auc        9.787857      35.249610  395.911031                 1.104288                1.778350          69.107899            2       True         20
28   NeuralNetFastAI_BAG_L1       0.646361   0.614142     roc_auc        2.099842       1.943866   68.738361                 2.099842                1.943866          68.738361            1       True         10
29    KNeighborsUnif_BAG_L1       0.578671   0.580084     roc_auc        0.133679       0.548662    0.084005                 0.133679                0.548662           0.084005            1       True          1
30    KNeighborsDist_BAG_L1       0.571540   0.579972     roc_auc        0.110390       0.606818    0.082456                 0.110390                0.606818           0.082456            1       True          2
	2	 = Optimal   num_stack_levels (Stacked Overfitting Occurred: False)
	934s	 = DyStack   runtime |	2666s	 = Remaining runtime
Starting main fit with num_stack_levels=2.
	For future fit calls on this dataset, you can skip DyStack to save time: `predictor.fit(..., dynamic_stacking=False, num_stack_levels=2)`
Beginning AutoGluon training ... Time limit = 2666s
AutoGluon will save models to "c:\studium\DAV Ausbildung\CADS\Immersion\ads3_exam\03_scripts\autogluon_diabetes_auc"
Train Data Rows:    13176
Train Data Columns: 2248
Label Column:       TARGET
Problem Type:       binary
Preprocessing data ...
Selected class <--> label mapping:  class 1 = 1, class 0 = 0
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
	Available Memory:                    2794.30 MB
	Train Data (Original)  Memory Usage: 225.98 MB (8.1% of available memory)
	Warning: Data size prior to feature transformation consumes 8.1% of available memory. Consider increasing memory or subsampling the data to avoid instability.
	Inferring data type of each feature based on column values. Set feature_metadata_in to manually specify special dtypes of the features.
	Stage 1 Generators:
		Fitting AsTypeFeatureGenerator...
			Note: Converting 1712 features to boolean dtype as they only contain 2 unique values.
	Stage 2 Generators:
		Fitting FillNaFeatureGenerator...
	Stage 3 Generators:
		Fitting IdentityFeatureGenerator...
	Stage 4 Generators:
		Fitting DropUniqueFeatureGenerator...
	Stage 5 Generators:
		Fitting DropDuplicatesFeatureGenerator...
	Useless Original Features (Count: 528): ['diag_1_110', 'diag_1_143', 'diag_1_147', 'diag_1_149', 'diag_1_160', 'diag_1_170', 'diag_1_173', 'diag_1_175', 'diag_1_187', 'diag_1_192', 'diag_1_194', 'diag_1_195', 'diag_1_200', 'diag_1_201', 'diag_1_207', 'diag_1_208', 'diag_1_210', 'diag_1_216', 'diag_1_219', 'diag_1_236', 'diag_1_246', 'diag_1_25021', 'diag_1_25032', 'diag_1_2505', 'diag_1_25053', 'diag_1_2509', 'diag_1_25091', 'diag_1_262', 'diag_1_266', 'diag_1_272', 'diag_1_301', 'diag_1_308', 'diag_1_318', 'diag_1_320', 'diag_1_322', 'diag_1_334', 'diag_1_335', 'diag_1_336', 'diag_1_337', 'diag_1_353', 'diag_1_36', 'diag_1_362', 'diag_1_363', 'diag_1_374', 'diag_1_377', 'diag_1_381', 'diag_1_382', 'diag_1_383', 'diag_1_385', 'diag_1_388', 'diag_1_39', 'diag_1_397', 'diag_1_412', 'diag_1_422', 'diag_1_445', 'diag_1_448', 'diag_1_457', 'diag_1_470', 'diag_1_477', 'diag_1_48', 'diag_1_49', 'diag_1_495', 'diag_1_501', 'diag_1_52', 'diag_1_521', 'diag_1_524', 'diag_1_526', 'diag_1_527', 'diag_1_528', 'diag_1_529', 'diag_1_542', 'diag_1_551', 'diag_1_565', 'diag_1_57', 'diag_1_573', 'diag_1_580', 'diag_1_582', 'diag_1_588', 'diag_1_603', 'diag_1_605', 'diag_1_61', 'diag_1_610', 'diag_1_632', 'diag_1_634', 'diag_1_643', 'diag_1_645', 'diag_1_646', 'diag_1_649', 'diag_1_653', 'diag_1_657', 'diag_1_665', 'diag_1_669', 'diag_1_671', 'diag_1_683', 'diag_1_684', 'diag_1_685', 'diag_1_690', 'diag_1_692', 'diag_1_693', 'diag_1_695', 'diag_1_7', 'diag_1_706', 'diag_1_709', 'diag_1_732', 'diag_1_734', 'diag_1_735', 'diag_1_745', 'diag_1_746', 'diag_1_747', 'diag_1_75', 'diag_1_791', 'diag_1_792', 'diag_1_795', 'diag_1_800', 'diag_1_810', 'diag_1_817', 'diag_1_831', 'diag_1_832', 'diag_1_835', 'diag_1_836', 'diag_1_84', 'diag_1_840', 'diag_1_845', 'diag_1_846', 'diag_1_848', 'diag_1_854', 'diag_1_862', 'diag_1_863', 'diag_1_868', 'diag_1_870', 'diag_1_871', 'diag_1_875', 'diag_1_879', 'diag_1_885', 'diag_1_886', 'diag_1_891', 'diag_1_892', 'diag_1_893', 'diag_1_897', 'diag_1_903', 'diag_1_906', 'diag_1_911', 'diag_1_914', 'diag_1_916', 'diag_1_921', 'diag_1_934', 'diag_1_936', 'diag_1_941', 'diag_1_955', 'diag_1_957', 'diag_1_963', 'diag_1_968', 'diag_1_976', 'diag_1_977', 'diag_1_98', 'diag_1_982', 'diag_1_983', 'diag_1_987', 'diag_1_E909', 'diag_1_V45', 'diag_1_V51', 'diag_1_V56', 'diag_1_V70', 'diag_2_111', 'diag_2_123', 'diag_2_131', 'diag_2_137', 'diag_2_145', 'diag_2_152', 'diag_2_163', 'diag_2_164', 'diag_2_173', 'diag_2_180', 'diag_2_192', 'diag_2_212', 'diag_2_217', 'diag_2_225', 'diag_2_228', 'diag_2_235', 'diag_2_239', 'diag_2_240', 'diag_2_259', 'diag_2_268', 'diag_2_269', 'diag_2_27', 'diag_2_271', 'diag_2_281', 'diag_2_297', 'diag_2_306', 'diag_2_31', 'diag_2_310', 'diag_2_316', 'diag_2_324', 'diag_2_338', 'diag_2_34', 'diag_2_347', 'diag_2_35', 'diag_2_351', 'diag_2_352', 'diag_2_355', 'diag_2_369', 'diag_2_373', 'diag_2_374', 'diag_2_376', 'diag_2_379', 'diag_2_382', 'diag_2_383', 'diag_2_389', 'diag_2_394', 'diag_2_395', 'diag_2_422', 'diag_2_430', 'diag_2_442', 'diag_2_464', 'diag_2_470', 'diag_2_484', 'diag_2_514', 'diag_2_520', 'diag_2_521', 'diag_2_522', 'diag_2_523', 'diag_2_527', 'diag_2_529', 'diag_2_534', 'diag_2_550', 'diag_2_557', 'diag_2_570', 'diag_2_579', 'diag_2_580', 'diag_2_588', 'diag_2_598', 'diag_2_604', 'diag_2_605', 'diag_2_615', 'diag_2_619', 'diag_2_622', 'diag_2_623', 'diag_2_627', 'diag_2_634', 'diag_2_641', 'diag_2_644', 'diag_2_645', 'diag_2_652', 'diag_2_658', 'diag_2_659', 'diag_2_661', 'diag_2_663', 'diag_2_664', 'diag_2_665', 'diag_2_674', 'diag_2_683', 'diag_2_686', 'diag_2_702', 'diag_2_703', 'diag_2_704', 'diag_2_705', 'diag_2_706', 'diag_2_713', 'diag_2_725', 'diag_2_734', 'diag_2_741', 'diag_2_747', 'diag_2_748', 'diag_2_751', 'diag_2_753', 'diag_2_758', 'diag_2_759', 'diag_2_795', 'diag_2_797', 'diag_2_800', 'diag_2_801', 'diag_2_814', 'diag_2_815', 'diag_2_831', 'diag_2_832', 'diag_2_837', 'diag_2_843', 'diag_2_861', 'diag_2_863', 'diag_2_867', 'diag_2_868', 'diag_2_870', 'diag_2_872', 'diag_2_880', 'diag_2_906', 'diag_2_910', 'diag_2_913', 'diag_2_915', 'diag_2_918', 'diag_2_922', 'diag_2_927', 'diag_2_933', 'diag_2_942', 'diag_2_944', 'diag_2_947', 'diag_2_952', 'diag_2_953', 'diag_2_962', 'diag_2_965', 'diag_2_967', 'diag_2_99', 'diag_2_991', 'diag_2_992', 'diag_2_E816', 'diag_2_E819', 'diag_2_E821', 'diag_2_E850', 'diag_2_E868', 'diag_2_E882', 'diag_2_E883', 'diag_2_E887', 'diag_2_E890', 'diag_2_E905', 'diag_2_E906', 'diag_2_E924', 'diag_2_E928', 'diag_2_E931', 'diag_2_E936', 'diag_2_E938', 'diag_2_E944', 'diag_2_E965', 'diag_2_E968', 'diag_2_E980', 'diag_2_V11', 'diag_2_V13', 'diag_2_V16', 'diag_2_V23', 'diag_2_V25', 'diag_2_V46', 'diag_2_V50', 'diag_2_V53', 'diag_2_V57', 'diag_2_V60', 'diag_2_V61', 'diag_2_V86', 'diag_3_115', 'diag_3_122', 'diag_3_123', 'diag_3_136', 'diag_3_139', 'diag_3_141', 'diag_3_150', 'diag_3_152', 'diag_3_161', 'diag_3_163', 'diag_3_17', 'diag_3_170', 'diag_3_171', 'diag_3_172', 'diag_3_182', 'diag_3_193', 'diag_3_214', 'diag_3_217', 'diag_3_223', 'diag_3_227', 'diag_3_233', 'diag_3_235', 'diag_3_240', 'diag_3_245', 'diag_3_25022', 'diag_3_2503', 'diag_3_25031', 'diag_3_258', 'diag_3_259', 'diag_3_265', 'diag_3_270', 'diag_3_271', 'diag_3_299', 'diag_3_314', 'diag_3_315', 'diag_3_318', 'diag_3_323', 'diag_3_35', 'diag_3_350', 'diag_3_359', 'diag_3_360', 'diag_3_361', 'diag_3_366', 'diag_3_370', 'diag_3_373', 'diag_3_374', 'diag_3_376', 'diag_3_379', 'diag_3_380', 'diag_3_381', 'diag_3_382', 'diag_3_383', 'diag_3_385', 'diag_3_387', 'diag_3_388', 'diag_3_395', 'diag_3_405', 'diag_3_417', 'diag_3_421', 'diag_3_430', 'diag_3_432', 'diag_3_436', 'diag_3_460', 'diag_3_463', 'diag_3_464', 'diag_3_47', 'diag_3_470', 'diag_3_472', 'diag_3_480', 'diag_3_49', 'diag_3_5', 'diag_3_500', 'diag_3_501', 'diag_3_506', 'diag_3_508', 'diag_3_514', 'diag_3_516', 'diag_3_523', 'diag_3_524', 'diag_3_525', 'diag_3_527', 'diag_3_534', 'diag_3_540', 'diag_3_542', 'diag_3_543', 'diag_3_556', 'diag_3_565', 'diag_3_566', 'diag_3_598', 'diag_3_608', 'diag_3_624', 'diag_3_641', 'diag_3_647', 'diag_3_655', 'diag_3_657', 'diag_3_665', 'diag_3_669', 'diag_3_674', 'diag_3_685', 'diag_3_694', 'diag_3_702', 'diag_3_706', 'diag_3_712', 'diag_3_734', 'diag_3_735', 'diag_3_741', 'diag_3_744', 'diag_3_746', 'diag_3_747', 'diag_3_750', 'diag_3_754', 'diag_3_755', 'diag_3_758', 'diag_3_759', 'diag_3_78', 'diag_3_801', 'diag_3_810', 'diag_3_814', 'diag_3_822', 'diag_3_826', 'diag_3_831', 'diag_3_834', 'diag_3_844', 'diag_3_845', 'diag_3_847', 'diag_3_851', 'diag_3_853', 'diag_3_854', 'diag_3_864', 'diag_3_866', 'diag_3_868', 'diag_3_871', 'diag_3_872', 'diag_3_876', 'diag_3_877', 'diag_3_881', 'diag_3_883', 'diag_3_890', 'diag_3_906', 'diag_3_910', 'diag_3_911', 'diag_3_918', 'diag_3_919', 'diag_3_930', 'diag_3_933', 'diag_3_943', 'diag_3_945', 'diag_3_951', 'diag_3_952', 'diag_3_953', 'diag_3_955', 'diag_3_956', 'diag_3_962', 'diag_3_965', 'diag_3_966', 'diag_3_967', 'diag_3_969', 'diag_3_970', 'diag_3_972', 'diag_3_987', 'diag_3_989', 'diag_3_E815', 'diag_3_E817', 'diag_3_E818', 'diag_3_E819', 'diag_3_E822', 'diag_3_E825', 'diag_3_E826', 'diag_3_E828', 'diag_3_E853', 'diag_3_E858', 'diag_3_E861', 'diag_3_E882', 'diag_3_E883', 'diag_3_E905', 'diag_3_E906', 'diag_3_E916', 'diag_3_E917', 'diag_3_E924', 'diag_3_E941', 'diag_3_E943', 'diag_3_E949', 'diag_3_E966', 'diag_3_E980', 'diag_3_V02', 'diag_3_V03', 'diag_3_V13', 'diag_3_V16', 'diag_3_V23', 'diag_3_V57', 'diag_3_V61', 'diag_3_V66', 'diag_3_V86']
		These features carry no predictive signal and should be manually investigated.
		This is typically a feature which has the same value for all rows.
		These features do not need to be present at inference time.
	Unused Original Features (Count: 626): ['admission_type_id_7', 'discharge_disposition_id_24', 'admission_source_id_8', 'payer_code_MP', 'medical_specialty_Gynecology', 'medical_specialty_Obstetrics', 'medical_specialty_PediatricsCriticalCare', 'medical_specialty_Podiatry', 'diag_1_11', 'diag_1_117', 'diag_1_133', 'diag_1_136', 'diag_1_141', 'diag_1_142', 'diag_1_146', 'diag_1_152', 'diag_1_158', 'diag_1_161', 'diag_1_179', 'diag_1_180', 'diag_1_184', 'diag_1_196', 'diag_1_199', 'diag_1_204', 'diag_1_205', 'diag_1_212', 'diag_1_215', 'diag_1_223', 'diag_1_229', 'diag_1_233', 'diag_1_237', 'diag_1_239', 'diag_1_242', 'diag_1_245', 'diag_1_2503', 'diag_1_25033', 'diag_1_25043', 'diag_1_25051', 'diag_1_251', 'diag_1_253', 'diag_1_255', 'diag_1_263', 'diag_1_273', 'diag_1_275', 'diag_1_282', 'diag_1_283', 'diag_1_288', 'diag_1_289', 'diag_1_3', 'diag_1_303', 'diag_1_304', 'diag_1_305', 'diag_1_307', 'diag_1_312', 'diag_1_314', 'diag_1_325', 'diag_1_34', 'diag_1_340', 'diag_1_342', 'diag_1_344', 'diag_1_345', 'diag_1_35', 'diag_1_352', 'diag_1_359', 'diag_1_360', 'diag_1_368', 'diag_1_370', 'diag_1_378', 'diag_1_379', 'diag_1_380', 'diag_1_384', 'diag_1_391', 'diag_1_394', 'diag_1_395', 'diag_1_396', 'diag_1_405', 'diag_1_41', 'diag_1_416', 'diag_1_42', 'diag_1_430', 'diag_1_442', 'diag_1_451', 'diag_1_452', 'diag_1_454', 'diag_1_456', 'diag_1_459', 'diag_1_462', 'diag_1_463', 'diag_1_464', 'diag_1_474', 'diag_1_478', 'diag_1_483', 'diag_1_492', 'diag_1_5', 'diag_1_508', 'diag_1_512', 'diag_1_514', 'diag_1_515', 'diag_1_516', 'diag_1_522', 'diag_1_533', 'diag_1_556', 'diag_1_568', 'diag_1_570', 'diag_1_579', 'diag_1_581', 'diag_1_591', 'diag_1_594', 'diag_1_598', 'diag_1_601', 'diag_1_602', 'diag_1_607', 'diag_1_611', 'diag_1_614', 'diag_1_617', 'diag_1_619', 'diag_1_620', 'diag_1_623', 'diag_1_627', 'diag_1_637', 'diag_1_641', 'diag_1_642', 'diag_1_644', 'diag_1_655', 'diag_1_659', 'diag_1_661', 'diag_1_663', 'diag_1_674', 'diag_1_686', 'diag_1_696', 'diag_1_70', 'diag_1_705', 'diag_1_708', 'diag_1_710', 'diag_1_714', 'diag_1_717', 'diag_1_720', 'diag_1_727', 'diag_1_736', 'diag_1_737', 'diag_1_753', 'diag_1_756', 'diag_1_759', 'diag_1_78', 'diag_1_788', 'diag_1_793', 'diag_1_796', 'diag_1_797', 'diag_1_801', 'diag_1_806', 'diag_1_816', 'diag_1_82', 'diag_1_826', 'diag_1_833', 'diag_1_843', 'diag_1_847', 'diag_1_853', 'diag_1_864', 'diag_1_865', 'diag_1_866', 'diag_1_867', 'diag_1_873', 'diag_1_880', 'diag_1_882', 'diag_1_890', 'diag_1_9', 'diag_1_913', 'diag_1_915', 'diag_1_922', 'diag_1_923', 'diag_1_924', 'diag_1_928', 'diag_1_933', 'diag_1_94', 'diag_1_952', 'diag_1_964', 'diag_1_965', 'diag_1_966', 'diag_1_970', 'diag_1_972', 'diag_1_986', 'diag_1_991', 'diag_1_992', 'diag_1_994', 'diag_1_999', 'diag_1_V26', 'diag_1_V55', 'diag_1_V60', 'diag_1_V71', 'diag_2_110', 'diag_2_114', 'diag_2_117', 'diag_2_135', 'diag_2_136', 'diag_2_140', 'diag_2_151', 'diag_2_155', 'diag_2_172', 'diag_2_179', 'diag_2_183', 'diag_2_188', 'diag_2_193', 'diag_2_200', 'diag_2_201', 'diag_2_208', 'diag_2_214', 'diag_2_215', 'diag_2_218', 'diag_2_220', 'diag_2_226', 'diag_2_227', 'diag_2_238', 'diag_2_246', 'diag_2_2501', 'diag_2_2502', 'diag_2_25022', 'diag_2_25033', 'diag_2_2505', 'diag_2_25052', 'diag_2_25053', 'diag_2_25083', 'diag_2_2509', 'diag_2_25091', 'diag_2_25093', 'diag_2_251', 'diag_2_252', 'diag_2_253', 'diag_2_258', 'diag_2_260', 'diag_2_261', 'diag_2_262', 'diag_2_273', 'diag_2_275', 'diag_2_279', 'diag_2_282', 'diag_2_283', 'diag_2_298', 'diag_2_301', 'diag_2_302', 'diag_2_307', 'diag_2_308', 'diag_2_317', 'diag_2_320', 'diag_2_322', 'diag_2_323', 'diag_2_327', 'diag_2_333', 'diag_2_344', 'diag_2_346', 'diag_2_353', 'diag_2_354', 'diag_2_360', 'diag_2_368', 'diag_2_372', 'diag_2_377', 'diag_2_380', 'diag_2_381', 'diag_2_386', 'diag_2_405', 'diag_2_421', 'diag_2_423', 'diag_2_436', 'diag_2_443', 'diag_2_447', 'diag_2_451', 'diag_2_452', 'diag_2_454', 'diag_2_455', 'diag_2_457', 'diag_2_459', 'diag_2_46', 'diag_2_460', 'diag_2_461', 'diag_2_462', 'diag_2_463', 'diag_2_472', 'diag_2_474', 'diag_2_485', 'diag_2_487', 'diag_2_494', 'diag_2_495', 'diag_2_500', 'diag_2_501', 'diag_2_508', 'diag_2_517', 'diag_2_528', 'diag_2_53', 'diag_2_533', 'diag_2_54', 'diag_2_540', 'diag_2_543', 'diag_2_555', 'diag_2_566', 'diag_2_586', 'diag_2_594', 'diag_2_600', 'diag_2_601', 'diag_2_602', 'diag_2_603', 'diag_2_611', 'diag_2_614', 'diag_2_617', 'diag_2_620', 'diag_2_621', 'diag_2_626', 'diag_2_646', 'diag_2_647', 'diag_2_649', 'diag_2_654', 'diag_2_681', 'diag_2_691', 'diag_2_692', 'diag_2_693', 'diag_2_696', 'diag_2_698', 'diag_2_701', 'diag_2_709', 'diag_2_712', 'diag_2_717', 'diag_2_718', 'diag_2_736', 'diag_2_738', 'diag_2_742', 'diag_2_746', 'diag_2_752', 'diag_2_755', 'diag_2_756', 'diag_2_783', 'diag_2_79', 'diag_2_791', 'diag_2_792', 'diag_2_793', 'diag_2_796', 'diag_2_802', 'diag_2_820', 'diag_2_825', 'diag_2_833', 'diag_2_836', 'diag_2_840', 'diag_2_842', 'diag_2_844', 'diag_2_847', 'diag_2_852', 'diag_2_853', 'diag_2_864', 'diag_2_865', 'diag_2_866', 'diag_2_871', 'diag_2_873', 'diag_2_88', 'diag_2_881', 'diag_2_882', 'diag_2_883', 'diag_2_891', 'diag_2_892', 'diag_2_9', 'diag_2_907', 'diag_2_911', 'diag_2_916', 'diag_2_919', 'diag_2_920', 'diag_2_921', 'diag_2_945', 'diag_2_948', 'diag_2_955', 'diag_2_958', 'diag_2_969', 'diag_2_972', 'diag_2_980', 'diag_2_989', 'diag_2_999', 'diag_2_E814', 'diag_2_E826', 'diag_2_E858', 'diag_2_E879', 'diag_2_E880', 'diag_2_E881', 'diag_2_E916', 'diag_2_E918', 'diag_2_E919', 'diag_2_E927', 'diag_2_E932', 'diag_2_E933', 'diag_2_E935', 'diag_2_E937', 'diag_2_E947', 'diag_2_E950', 'diag_2_V02', 'diag_2_V03', 'diag_2_V09', 'diag_2_V10', 'diag_2_V14', 'diag_2_V17', 'diag_2_V44', 'diag_2_V63', 'diag_2_V65', 'diag_2_V70', 'diag_3_117', 'diag_3_131', 'diag_3_135', 'diag_3_14', 'diag_3_151', 'diag_3_153', 'diag_3_154', 'diag_3_156', 'diag_3_157', 'diag_3_158', 'diag_3_173', 'diag_3_179', 'diag_3_180', 'diag_3_183', 'diag_3_188', 'diag_3_189', 'diag_3_200', 'diag_3_201', 'diag_3_202', 'diag_3_208', 'diag_3_220', 'diag_3_228', 'diag_3_239', 'diag_3_242', 'diag_3_2501', 'diag_3_25011', 'diag_3_25013', 'diag_3_2502', 'diag_3_25083', 'diag_3_2509', 'diag_3_25091', 'diag_3_25093', 'diag_3_251', 'diag_3_252', 'diag_3_256', 'diag_3_262', 'diag_3_266', 'diag_3_268', 'diag_3_27', 'diag_3_273', 'diag_3_277', 'diag_3_279', 'diag_3_281', 'diag_3_282', 'diag_3_283', 'diag_3_289', 'diag_3_293', 'diag_3_297', 'diag_3_298', 'diag_3_306', 'diag_3_312', 'diag_3_313', 'diag_3_334', 'diag_3_335', 'diag_3_336', 'diag_3_338', 'diag_3_341', 'diag_3_343', 'diag_3_347', 'diag_3_348', 'diag_3_351', 'diag_3_354', 'diag_3_355', 'diag_3_356', 'diag_3_358', 'diag_3_36544', 'diag_3_369', 'diag_3_378', 'diag_3_384', 'diag_3_386', 'diag_3_394', 'diag_3_398', 'diag_3_423', 'diag_3_431', 'diag_3_434', 'diag_3_446', 'diag_3_451', 'diag_3_454', 'diag_3_462', 'diag_3_477', 'diag_3_478', 'diag_3_494', 'diag_3_495', 'diag_3_510', 'diag_3_512', 'diag_3_517', 'diag_3_522', 'diag_3_529', 'diag_3_53', 'diag_3_532', 'diag_3_533', 'diag_3_537', 'diag_3_54', 'diag_3_550', 'diag_3_555', 'diag_3_57', 'diag_3_570', 'diag_3_572', 'diag_3_575', 'diag_3_579', 'diag_3_580', 'diag_3_582', 'diag_3_586', 'diag_3_594', 'diag_3_602', 'diag_3_604', 'diag_3_605', 'diag_3_607', 'diag_3_614', 'diag_3_619', 'diag_3_623', 'diag_3_627', 'diag_3_642', 'diag_3_644', 'diag_3_646', 'diag_3_653', 'diag_3_659', 'diag_3_661', 'diag_3_663', 'diag_3_664', 'diag_3_670', 'diag_3_686', 'diag_3_690', 'diag_3_692', 'diag_3_693', 'diag_3_695', 'diag_3_697', 'diag_3_698', 'diag_3_701', 'diag_3_703', 'diag_3_704', 'diag_3_705', 'diag_3_708', 'diag_3_709', 'diag_3_710', 'diag_3_713', 'diag_3_717', 'diag_3_718', 'diag_3_721', 'diag_3_723', 'diag_3_725', 'diag_3_726', 'diag_3_736', 'diag_3_737', 'diag_3_745', 'diag_3_751', 'diag_3_752', 'diag_3_79', 'diag_3_792', 'diag_3_793', 'diag_3_795', 'diag_3_800', 'diag_3_805', 'diag_3_808', 'diag_3_816', 'diag_3_820', 'diag_3_823', 'diag_3_825', 'diag_3_836', 'diag_3_838', 'diag_3_842', 'diag_3_850', 'diag_3_852', 'diag_3_860', 'diag_3_862', 'diag_3_865', 'diag_3_873', 'diag_3_875', 'diag_3_880', 'diag_3_891', 'diag_3_9', 'diag_3_905', 'diag_3_907', 'diag_3_908', 'diag_3_912', 'diag_3_913', 'diag_3_917', 'diag_3_920', 'diag_3_921', 'diag_3_922', 'diag_3_924', 'diag_3_928', 'diag_3_934', 'diag_3_948', 'diag_3_958', 'diag_3_971', 'diag_3_991', 'diag_3_999', 'diag_3_E812', 'diag_3_E813', 'diag_3_E850', 'diag_3_E852', 'diag_3_E870', 'diag_3_E886', 'diag_3_E892', 'diag_3_E894', 'diag_3_E900', 'diag_3_E912', 'diag_3_E919', 'diag_3_E920', 'diag_3_E922', 'diag_3_E928', 'diag_3_E930', 'diag_3_E931', 'diag_3_E932', 'diag_3_E933', 'diag_3_E934', 'diag_3_E936', 'diag_3_E937', 'diag_3_E938', 'diag_3_E946', 'diag_3_E950', 'diag_3_E956', 'diag_3_V06', 'diag_3_V14', 'diag_3_V17', 'diag_3_V25', 'diag_3_V44', 'diag_3_V55', 'diag_3_V62', 'diag_3_V70', 'diag_3_V72', 'nateglinide_Up', 'chlorpropamide_Steady', 'tolbutamide_Steady', 'miglitol_Steady', 'group_diag_1_na', 'group_diag_2_na', 'group_diag_3_na']
		These features were not used to generate any of the output features. Add a feature generator compatible with these features to utilize them.
		Features can also be unused if they carry very little information, such as being categorical but having almost entirely unique values or being duplicates of other features.
		These features do not need to be present at inference time.
		('float', []) : 626 | ['admission_type_id_7', 'discharge_disposition_id_24', 'admission_source_id_8', 'payer_code_MP', 'medical_specialty_Gynecology', ...]
	Types of features in original data (raw dtype, special dtypes):
		('float', []) : 1094 | ['time_in_hospital', 'num_lab_procedures', 'num_procedures', 'num_medications', 'number_outpatient', ...]
	Types of features in processed data (raw dtype, special dtypes):
		('float', [])     :    8 | ['time_in_hospital', 'num_lab_procedures', 'num_procedures', 'num_medications', 'number_outpatient', ...]
		('int', ['bool']) : 1086 | ['race_AfricanAmerican', 'race_Asian', 'race_Caucasian', 'race_Hispanic', 'race_Other', ...]
	16.2s = Fit runtime
	1094 features in original data used to generate 1094 features in processed data.
	Train Data (Processed) Memory Usage: 14.45 MB (0.5% of available memory)
Data preprocessing and feature engineering runtime = 16.94s ...
AutoGluon will gauge predictive performance using evaluation metric: 'roc_auc'
	This metric expects predicted probabilities rather than predicted class labels, so you'll need to use predict_proba() instead of predict()
	To change this, specify the eval_metric parameter of Predictor()
Large model count detected (112 configs) ... Only displaying the first 3 models of each family. To see all, set `verbosity=3`.
User-specified model hyperparameters to be fit:
{
	'NN_TORCH': [{}, {'activation': 'elu', 'dropout_prob': 0.10077639529843717, 'hidden_size': 108, 'learning_rate': 0.002735937344002146, 'num_layers': 4, 'use_batchnorm': True, 'weight_decay': 1.356433327634438e-12, 'ag_args': {'name_suffix': '_r79', 'priority': -2}}, {'activation': 'elu', 'dropout_prob': 0.11897478034205347, 'hidden_size': 213, 'learning_rate': 0.0010474382260641949, 'num_layers': 4, 'use_batchnorm': False, 'weight_decay': 5.594471067786272e-10, 'ag_args': {'name_suffix': '_r22', 'priority': -7}}],
	'GBM': [{'extra_trees': True, 'ag_args': {'name_suffix': 'XT'}}, {}, {'learning_rate': 0.03, 'num_leaves': 128, 'feature_fraction': 0.9, 'min_data_in_leaf': 3, 'ag_args': {'name_suffix': 'Large', 'priority': 0, 'hyperparameter_tune_kwargs': None}}],
	'CAT': [{}, {'depth': 6, 'grow_policy': 'SymmetricTree', 'l2_leaf_reg': 2.1542798306067823, 'learning_rate': 0.06864209415792857, 'max_ctr_complexity': 4, 'one_hot_max_size': 10, 'ag_args': {'name_suffix': '_r177', 'priority': -1}}, {'depth': 8, 'grow_policy': 'Depthwise', 'l2_leaf_reg': 2.7997999596449104, 'learning_rate': 0.031375015734637225, 'max_ctr_complexity': 2, 'one_hot_max_size': 3, 'ag_args': {'name_suffix': '_r9', 'priority': -5}}],
	'XGB': [{}, {'colsample_bytree': 0.6917311125174739, 'enable_categorical': False, 'learning_rate': 0.018063876087523967, 'max_depth': 10, 'min_child_weight': 0.6028633586934382, 'ag_args': {'name_suffix': '_r33', 'priority': -8}}, {'colsample_bytree': 0.6628423832084077, 'enable_categorical': False, 'learning_rate': 0.08775715546881824, 'max_depth': 5, 'min_child_weight': 0.6294123374222513, 'ag_args': {'name_suffix': '_r89', 'priority': -16}}],
	'FASTAI': [{}, {'bs': 256, 'emb_drop': 0.5411770367537934, 'epochs': 43, 'layers': [800, 400], 'lr': 0.01519848858318159, 'ps': 0.23782946566604385, 'ag_args': {'name_suffix': '_r191', 'priority': -4}}, {'bs': 2048, 'emb_drop': 0.05070411322605811, 'epochs': 29, 'layers': [200, 100], 'lr': 0.08974235041576624, 'ps': 0.10393466140748028, 'ag_args': {'name_suffix': '_r102', 'priority': -11}}],
	'RF': [{'criterion': 'gini', 'ag_args': {'name_suffix': 'Gini', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'entropy', 'ag_args': {'name_suffix': 'Entr', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'squared_error', 'ag_args': {'name_suffix': 'MSE', 'problem_types': ['regression', 'quantile']}}],
	'XT': [{'criterion': 'gini', 'ag_args': {'name_suffix': 'Gini', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'entropy', 'ag_args': {'name_suffix': 'Entr', 'problem_types': ['binary', 'multiclass']}}, {'criterion': 'squared_error', 'ag_args': {'name_suffix': 'MSE', 'problem_types': ['regression', 'quantile']}}],
	'KNN': [{'weights': 'uniform', 'ag_args': {'name_suffix': 'Unif'}}, {'weights': 'distance', 'ag_args': {'name_suffix': 'Dist'}}],
}
AutoGluon will fit 3 stack levels (L1 to L3) ...
Fitting 110 L1 models, fit_strategy="sequential" ...
Fitting model: KNeighborsUnif_BAG_L1 ... Training model for up to 1177.27s of the 2649.49s of remaining time.
	0.576	 = Validation score   (roc_auc)
	0.1s	 = Training   runtime
	0.73s	 = Validation runtime
Fitting model: KNeighborsDist_BAG_L1 ... Training model for up to 1172.88s of the 2645.11s of remaining time.
	0.5757	 = Validation score   (roc_auc)
	0.1s	 = Training   runtime
	0.66s	 = Validation runtime
Fitting model: LightGBMXT_BAG_L1 ... Training model for up to 1171.89s of the 2644.11s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 12.73% memory usage per fold, 50.91%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=12.73%)
	0.6945	 = Validation score   (roc_auc)
	13.44s	 = Training   runtime
	0.46s	 = Validation runtime
Fitting model: LightGBM_BAG_L1 ... Training model for up to 1152.53s of the 2624.75s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 16.03% memory usage per fold, 64.12%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=16.03%)
	0.6954	 = Validation score   (roc_auc)
	13.41s	 = Training   runtime
	0.43s	 = Validation runtime
Fitting model: RandomForestGini_BAG_L1 ... Training model for up to 1133.70s of the 2605.93s of remaining time.
	0.6826	 = Validation score   (roc_auc)
	10.53s	 = Training   runtime
	9.0s	 = Validation runtime
Fitting model: RandomForestEntr_BAG_L1 ... Training model for up to 1113.15s of the 2585.38s of remaining time.
	0.6845	 = Validation score   (roc_auc)
	12.15s	 = Training   runtime
	8.36s	 = Validation runtime
Fitting model: CatBoost_BAG_L1 ... Training model for up to 1091.55s of the 2563.77s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 1 folds in parallel instead (Estimated 60.11% memory usage per fold, 60.11%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (1 workers, per: cpus=6, gpus=0, memory=60.11%)
		Switching to pseudo sequential ParallelFoldFittingStrategy to avoid Python memory leakage.
		Overrule this behavior by setting fold_fitting_strategy to 'sequential_local' in ag_args_ensemble when when calling `predictor.fit`
	0.6983	 = Validation score   (roc_auc)
	132.55s	 = Training   runtime
	1.37s	 = Validation runtime
Fitting model: ExtraTreesGini_BAG_L1 ... Training model for up to 953.60s of the 2425.83s of remaining time.
	0.6758	 = Validation score   (roc_auc)
	11.45s	 = Training   runtime
	7.15s	 = Validation runtime
Fitting model: ExtraTreesEntr_BAG_L1 ... Training model for up to 934.13s of the 2406.35s of remaining time.
	0.6806	 = Validation score   (roc_auc)
	12.1s	 = Training   runtime
	7.62s	 = Validation runtime
Fitting model: NeuralNetFastAI_BAG_L1 ... Training model for up to 913.46s of the 2385.68s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=6.16%)
	0.6231	 = Validation score   (roc_auc)
	81.84s	 = Training   runtime
	1.73s	 = Validation runtime
Fitting model: XGBoost_BAG_L1 ... Training model for up to 823.62s of the 2295.85s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=7.15%)
	0.6923	 = Validation score   (roc_auc)
	47.71s	 = Training   runtime
	0.61s	 = Validation runtime
Fitting model: NeuralNetTorch_BAG_L1 ... Training model for up to 767.82s of the 2240.04s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.12%)
	0.6888	 = Validation score   (roc_auc)
	56.36s	 = Training   runtime
	0.85s	 = Validation runtime
Fitting model: LightGBMLarge_BAG_L1 ... Training model for up to 703.63s of the 2175.85s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 13.32% memory usage per fold, 53.28%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=13.32%)
	0.6866	 = Validation score   (roc_auc)
	29.96s	 = Training   runtime
	0.67s	 = Validation runtime
Fitting model: CatBoost_r177_BAG_L1 ... Training model for up to 667.72s of the 2139.95s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 19.65% memory usage per fold, 78.61%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=19.65%)
	0.6984	 = Validation score   (roc_auc)
	44.74s	 = Training   runtime
	1.76s	 = Validation runtime
Fitting model: NeuralNetTorch_r79_BAG_L1 ... Training model for up to 616.62s of the 2088.85s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.20%)
	0.6536	 = Validation score   (roc_auc)
	62.85s	 = Training   runtime
	0.89s	 = Validation runtime
Fitting model: LightGBM_r131_BAG_L1 ... Training model for up to 545.57s of the 2017.79s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=6.53%)
	0.6946	 = Validation score   (roc_auc)
	10.85s	 = Training   runtime
	1.29s	 = Validation runtime
Fitting model: NeuralNetFastAI_r191_BAG_L1 ... Training model for up to 526.30s of the 1998.53s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.32%)
	0.6101	 = Validation score   (roc_auc)
	140.0s	 = Training   runtime
	2.18s	 = Validation runtime
Fitting model: CatBoost_r9_BAG_L1 ... Training model for up to 379.89s of the 1852.11s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 34.23% memory usage per fold, 68.47%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=34.23%)
	0.6971	 = Validation score   (roc_auc)
	184.69s	 = Training   runtime
	1.67s	 = Validation runtime
Fitting model: LightGBM_r96_BAG_L1 ... Training model for up to 190.17s of the 1662.40s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.88%)
	0.6976	 = Validation score   (roc_auc)
	9.5s	 = Training   runtime
	1.49s	 = Validation runtime
Fitting model: NeuralNetTorch_r22_BAG_L1 ... Training model for up to 174.02s of the 1646.24s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.28%)
	0.6807	 = Validation score   (roc_auc)
	71.46s	 = Training   runtime
	0.78s	 = Validation runtime
Fitting model: XGBoost_r33_BAG_L1 ... Training model for up to 96.13s of the 1568.35s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 1 folds in parallel instead (Estimated 49.97% memory usage per fold, 49.97%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (1 workers, per: cpus=6, gpus=0, memory=49.97%)
		Switching to pseudo sequential ParallelFoldFittingStrategy to avoid Python memory leakage.
		Overrule this behavior by setting fold_fitting_strategy to 'sequential_local' in ag_args_ensemble when when calling `predictor.fit`
	0.6887	 = Validation score   (roc_auc)
	103.25s	 = Training   runtime
	0.8s	 = Validation runtime
Fitting model: WeightedEnsemble_L2 ... Training model for up to 360.00s of the 1460.75s of remaining time.
	Ensemble Weights: {'CatBoost_r177_BAG_L1': 0.28, 'LightGBM_r96_BAG_L1': 0.24, 'NeuralNetTorch_BAG_L1': 0.12, 'LightGBM_BAG_L1': 0.08, 'RandomForestEntr_BAG_L1': 0.08, 'CatBoost_BAG_L1': 0.04, 'ExtraTreesEntr_BAG_L1': 0.04, 'NeuralNetTorch_r79_BAG_L1': 0.04, 'CatBoost_r9_BAG_L1': 0.04, 'NeuralNetTorch_r22_BAG_L1': 0.04}
	0.7056	 = Validation score   (roc_auc)
	1.47s	 = Training   runtime
	0.0s	 = Validation runtime
Fitting 108 L2 models, fit_strategy="sequential" ...
Fitting model: LightGBMXT_BAG_L2 ... Training model for up to 972.56s of the 1459.01s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=4.34%)
	0.7006	 = Validation score   (roc_auc)
	7.33s	 = Training   runtime
	0.83s	 = Validation runtime
Fitting model: LightGBM_BAG_L2 ... Training model for up to 956.40s of the 1442.84s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=4.48%)
	0.6984	 = Validation score   (roc_auc)
	8.17s	 = Training   runtime
	0.62s	 = Validation runtime
Fitting model: RandomForestGini_BAG_L2 ... Training model for up to 938.13s of the 1424.57s of remaining time.
	0.6964	 = Validation score   (roc_auc)
	8.81s	 = Training   runtime
	7.19s	 = Validation runtime
Fitting model: RandomForestEntr_BAG_L2 ... Training model for up to 921.08s of the 1407.52s of remaining time.
	0.6968	 = Validation score   (roc_auc)
	8.43s	 = Training   runtime
	6.65s	 = Validation runtime
Fitting model: CatBoost_BAG_L2 ... Training model for up to 905.16s of the 1391.60s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 23.25% memory usage per fold, 46.49%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=23.25%)
	0.7037	 = Validation score   (roc_auc)
	53.53s	 = Training   runtime
	1.89s	 = Validation runtime
Fitting model: ExtraTreesGini_BAG_L2 ... Training model for up to 846.98s of the 1333.42s of remaining time.
	0.696	 = Validation score   (roc_auc)
	12.57s	 = Training   runtime
	7.44s	 = Validation runtime
Fitting model: ExtraTreesEntr_BAG_L2 ... Training model for up to 825.88s of the 1312.32s of remaining time.
	0.6936	 = Validation score   (roc_auc)
	11.61s	 = Training   runtime
	7.31s	 = Validation runtime
Fitting model: NeuralNetFastAI_BAG_L2 ... Training model for up to 805.91s of the 1292.35s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=3.15%)
	0.6164	 = Validation score   (roc_auc)
	71.66s	 = Training   runtime
	1.65s	 = Validation runtime
Fitting model: XGBoost_BAG_L2 ... Training model for up to 725.05s of the 1211.50s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=8.40%)
	0.6956	 = Validation score   (roc_auc)
	34.53s	 = Training   runtime
	0.67s	 = Validation runtime
Fitting model: NeuralNetTorch_BAG_L2 ... Training model for up to 682.63s of the 1169.08s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.36%)
	0.6779	 = Validation score   (roc_auc)
	64.64s	 = Training   runtime
	0.71s	 = Validation runtime
Fitting model: LightGBMLarge_BAG_L2 ... Training model for up to 610.36s of the 1096.80s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 14.87% memory usage per fold, 59.50%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=14.87%)
	0.6941	 = Validation score   (roc_auc)
	29.61s	 = Training   runtime
	0.51s	 = Validation runtime
Fitting model: CatBoost_r177_BAG_L2 ... Training model for up to 576.50s of the 1062.94s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 21.55% memory usage per fold, 43.10%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=21.55%)
	0.7034	 = Validation score   (roc_auc)
	54.54s	 = Training   runtime
	1.67s	 = Validation runtime
Fitting model: NeuralNetTorch_r79_BAG_L2 ... Training model for up to 517.10s of the 1003.54s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.36%)
	0.675	 = Validation score   (roc_auc)
	50.12s	 = Training   runtime
	1.12s	 = Validation runtime
Fitting model: LightGBM_r131_BAG_L2 ... Training model for up to 457.13s of the 943.57s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=7.32%)
	0.6975	 = Validation score   (roc_auc)
	10.98s	 = Training   runtime
	0.82s	 = Validation runtime
Fitting model: NeuralNetFastAI_r191_BAG_L2 ... Training model for up to 438.90s of the 925.34s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.64%)
	0.6253	 = Validation score   (roc_auc)
	153.24s	 = Training   runtime
	2.7s	 = Validation runtime
Fitting model: CatBoost_r9_BAG_L2 ... Training model for up to 278.29s of the 764.73s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 38.14% memory usage per fold, 76.28%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=38.14%)
	0.7003	 = Validation score   (roc_auc)
	155.25s	 = Training   runtime
	1.9s	 = Validation runtime
Fitting model: LightGBM_r96_BAG_L2 ... Training model for up to 115.90s of the 602.34s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.65%)
	0.7019	 = Validation score   (roc_auc)
	6.53s	 = Training   runtime
	0.85s	 = Validation runtime
Fitting model: NeuralNetTorch_r22_BAG_L2 ... Training model for up to 101.32s of the 587.76s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.22%)
	0.7141	 = Validation score   (roc_auc)
	88.44s	 = Training   runtime
	1.32s	 = Validation runtime
Fitting model: XGBoost_r33_BAG_L2 ... Training model for up to 5.61s of the 492.05s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 1 folds in parallel instead (Estimated 56.21% memory usage per fold, 56.21%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (1 workers, per: cpus=6, gpus=0, memory=56.21%)
		Switching to pseudo sequential ParallelFoldFittingStrategy to avoid Python memory leakage.
		Overrule this behavior by setting fold_fitting_strategy to 'sequential_local' in ag_args_ensemble when when calling `predictor.fit`
	0.6385	 = Validation score   (roc_auc)
	36.6s	 = Training   runtime
	0.67s	 = Validation runtime
Fitting model: WeightedEnsemble_L3 ... Training model for up to 360.00s of the 450.22s of remaining time.
	Ensemble Weights: {'CatBoost_BAG_L2': 0.5, 'NeuralNetTorch_r22_BAG_L2': 0.5}
	0.7263	 = Validation score   (roc_auc)
	1.44s	 = Training   runtime
	0.0s	 = Validation runtime
Fitting 108 L3 models, fit_strategy="sequential" ...
Fitting model: LightGBMXT_BAG_L3 ... Training model for up to 448.74s of the 448.53s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=4.14%)
	0.7204	 = Validation score   (roc_auc)
	8.83s	 = Training   runtime
	1.09s	 = Validation runtime
Fitting model: LightGBM_BAG_L3 ... Training model for up to 427.50s of the 427.29s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=4.21%)
	0.722	 = Validation score   (roc_auc)
	8.75s	 = Training   runtime
	0.64s	 = Validation runtime
Fitting model: RandomForestGini_BAG_L3 ... Training model for up to 408.96s of the 408.75s of remaining time.
	0.7102	 = Validation score   (roc_auc)
	8.57s	 = Training   runtime
	6.13s	 = Validation runtime
Fitting model: RandomForestEntr_BAG_L3 ... Training model for up to 393.11s of the 392.90s of remaining time.
	0.7115	 = Validation score   (roc_auc)
	8.03s	 = Training   runtime
	5.8s	 = Validation runtime
Fitting model: CatBoost_BAG_L3 ... Training model for up to 378.52s of the 378.30s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 21.74% memory usage per fold, 43.48%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=21.74%)
	0.7254	 = Validation score   (roc_auc)
	43.53s	 = Training   runtime
	1.22s	 = Validation runtime
Fitting model: ExtraTreesGini_BAG_L3 ... Training model for up to 331.74s of the 331.52s of remaining time.
	0.7063	 = Validation score   (roc_auc)
	9.12s	 = Training   runtime
	5.58s	 = Validation runtime
Fitting model: ExtraTreesEntr_BAG_L3 ... Training model for up to 316.34s of the 316.12s of remaining time.
	0.708	 = Validation score   (roc_auc)
	9.42s	 = Training   runtime
	5.85s	 = Validation runtime
Fitting model: NeuralNetFastAI_BAG_L3 ... Training model for up to 300.39s of the 300.17s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=2.53%)
	0.6154	 = Validation score   (roc_auc)
	70.78s	 = Training   runtime
	1.62s	 = Validation runtime
Fitting model: XGBoost_BAG_L3 ... Training model for up to 221.52s of the 221.31s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=9.09%)
	0.7217	 = Validation score   (roc_auc)
	36.58s	 = Training   runtime
	0.87s	 = Validation runtime
Fitting model: NeuralNetTorch_BAG_L3 ... Training model for up to 177.98s of the 177.76s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.47%)
	0.7154	 = Validation score   (roc_auc)
	58.22s	 = Training   runtime
	0.86s	 = Validation runtime
Fitting model: LightGBMLarge_BAG_L3 ... Training model for up to 110.94s of the 110.72s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 4 folds in parallel instead (Estimated 15.97% memory usage per fold, 63.90%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (4 workers, per: cpus=3, gpus=0, memory=15.97%)
	0.7169	 = Validation score   (roc_auc)
	33.2s	 = Training   runtime
	0.54s	 = Validation runtime
Fitting model: CatBoost_r177_BAG_L3 ... Training model for up to 73.04s of the 72.83s of remaining time.
	Memory not enough to fit 8 folds in parallel. Will train 2 folds in parallel instead (Estimated 22.37% memory usage per fold, 44.74%/80.00% total).
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (2 workers, per: cpus=6, gpus=0, memory=22.37%)
	0.7256	 = Validation score   (roc_auc)
	54.74s	 = Training   runtime
	1.78s	 = Validation runtime
Fitting model: NeuralNetTorch_r79_BAG_L3 ... Training model for up to 12.85s of the 12.64s of remaining time.
	Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelLocalFoldFittingStrategy (8 workers, per: cpus=1, gpus=0, memory=1.62%)
	0.682	 = Validation score   (roc_auc)
	16.25s	 = Training   runtime
	0.98s	 = Validation runtime
Fitting model: WeightedEnsemble_L4 ... Training model for up to 360.00s of the -12.66s of remaining time.
	Ensemble Weights: {'NeuralNetTorch_r22_BAG_L2': 0.24, 'CatBoost_BAG_L2': 0.2, 'CatBoost_r177_BAG_L3': 0.16, 'LightGBM_BAG_L3': 0.12, 'XGBoost_BAG_L3': 0.12, 'XGBoost_BAG_L1': 0.08, 'NeuralNetTorch_BAG_L1': 0.04, 'CatBoost_BAG_L3': 0.04}
	0.7276	 = Validation score   (roc_auc)
	2.06s	 = Training   runtime
	0.0s	 = Validation runtime
AutoGluon training complete, total runtime = 2681.28s ... Best model: WeightedEnsemble_L4 | Estimated inference throughput: 35.8 rows/s (1647 batch size)
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("c:\studium\DAV Ausbildung\CADS\Immersion\ads3_exam\03_scripts\autogluon_diabetes_auc")

Elapsed time (sec): 3615.57

# View leaderboard
leaderboard = AG.leaderboard(Xy_val, silent=True)
leaderboard.head(15)

# Evaluate on validation data
performance = AG.evaluate(Xy_val)

# Extract and print the validation AUC score
auc_AG = performance['roc_auc']
print(f'The validation AUC of the best model from AutoGluon is: {auc_AG:.6f}')

The validation AUC of the best model from AutoGluon is: 0.700962

# Choose the best ensemble model (here the random forest)
WE = leaderboard.iloc[6]['model']

performance = AG.evaluate(Xy_val, model=WE)

# Extract and print the validation AUC score
auc_WE = performance['roc_auc']
print(f'The validation AUC of the best model from AutoGluon is: {auc_WE:.6f}')

The validation AUC of the best model from AutoGluon is: 0.704813

# Feature scaling using StandardScaler based on training data distributions (6.1)
X_test = pd.DataFrame(scaler.transform(X_test), columns = feature_names)

# Make preditions for test data
y_test_CB1 = CB1.predict_proba(X_raw_test)
y_test_LR1 = LR1.predict(X_pre_test)

y_test_CB4 = CB4.predict_proba(X_test)
y_test_CB5 = CB5.predict_proba(X_test)
y_test_CB6 = CB6.predict_proba(X_test)
y_test_LGB = LGB.predict_proba(X_test)
y_test_XGB = XGB.predict_proba(X_test)

y_test_ANN = ANN.predict(X2_test_ANN)
y_test_LR2 = LR2.predict(X2_test)
y_test_CB7 = CB7.predict_proba(X2_test)
y_test_CB8 = CB8.predict_proba(X2_all_test)

y_test_WE = AG.predict_proba(X_test, model=WE)

224/224 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step

# Calulate model AUC based on test data 

auc_CB1_test = roc_auc_score(y_test, y_test_CB1[:, 1])
auc_LR1_test = roc_auc_score(y_test, y_test_LR1)

# Calculate the further model test AUC
auc_CB4_test = roc_auc_score(y_test, y_test_CB4[:, 1])
auc_CB5_test = roc_auc_score(y_test, y_test_CB5[:, 1])
auc_CB6_test = roc_auc_score(y_test, y_test_CB6[:, 1])
auc_LGB_test = roc_auc_score(y_test, y_test_LGB[:, 1])
auc_XGB_test = roc_auc_score(y_test, y_test_XGB[:, 1])

auc_ANN_test = roc_auc_score(y2_test, y_test_ANN[:, 0])
auc_LR2_test = roc_auc_score(y2_test, y_test_LR2)
auc_CB7_test = roc_auc_score(y2_test, y_test_CB7[:, 1])
auc_CB8_test = roc_auc_score(y2_test, y_test_CB8[:, 1])

auc_WE_test = roc_auc_score(y_test, y_test_WE[1])

# Data structures for model names, val/test-sample and AUC
mdict = {'Model name': [], 'Set': [], 'AUC': []}

# List of models and their AUC scores on validation and test sets
models_auc = [
    ("CB1_quick",           auc_CB1, auc_CB1_test),
    ("LR1_select",          auc_LR1, auc_LR1_test),
    ("CB4_OHE",             auc_CB4, auc_CB4_test),
    ("CB5_subsample",       auc_CB5, auc_CB5_test),
    ("CB6_tuned",           auc_CB6, auc_CB6_test),
    ("LGB_tuned",           auc_LGB, auc_LGB_test),
    ("XGB_tuned",           auc_XGB, auc_XGB_test),
    ("ANN_tuned",           auc_ANN, auc_ANN_test),
    ("LR2_with_embeddings", auc_LR2, auc_LR2_test),
    ("CB7_with_embeddings", auc_CB7, auc_CB7_test),
    ("CB8_quick_with_embeddings", auc_CB8, auc_CB8_test),
    ("WeightedEnsemble",    auc_WE, auc_WE_test),
]

# Populate the dictionary by looping over the models and their AUC scores
for model_name, auc_val, auc_test in models_auc:
    mdict['Model name'].extend([model_name] * 2)  # Model name twice for Val. and Test
    mdict['Set'].extend(["Val.", "Test"])
    mdict['AUC'].extend([auc_val, auc_test])

df_eval = pd.DataFrame(mdict)
plt.title("Model evaluation (AUC):")
sns.barplot(data=df_eval, x="AUC", y="Model name",hue="Set")
plt.xlim(0.50, 0.76)
plt.show()

# Core of scikit-plot function cumulative_gain_curve (necessary, since scypi release 1.12.0 broke compatibility with scikit-plot version 0.3.7.
# Source: https://github.com/reiinakano/scikit-plot/blob/26007fbf9f05e915bd0f6acb86850b01b00944cf/scikitplot/helpers.py
def cumulative_gain_curve1(y_true, y_score):
    """This binary classification function generates the points necessary to plot the Cumulative Gain"""
    y_true, y_score = np.asarray(y_true), np.asarray(y_score)
    # make y_true a boolean vector
    y_true = (y_true == 1)
    sorted_indices = np.argsort(y_score)[::-1]
    y_true = y_true[sorted_indices]
    gains = np.cumsum(y_true)
    percentages = np.arange(start=1, stop=len(y_true) + 1)
    gains = gains / float(np.sum(y_true))
    percentages = percentages / float(len(y_true))
    gains = np.insert(gains, 0, [0])
    percentages = np.insert(percentages, 0, [0])
    return percentages, gains

# Calculate cumulative gains for lift chart 
def calculate_gains_and_percentages(model_predictions, y_true):
    """Calculate gains and percentages for a given model's predictions."""
    percentages, gains = cumulative_gain_curve1(y_true, model_predictions)
    gains_adjusted = gains[1:] / percentages[1:]  # Adjust gains starting from the second element
    return percentages[1:], gains_adjusted

# Select models and prepare true and predicted values
y_true = np.array(np.ravel(y_test))
model_predictions = {
    'CB1_quick': np.array(y_test_CB1[:, 1]),
    'CB5_subsample': np.array(y_test_CB5[:, 1]),
    'LGB_tuned': np.array(y_test_LGB[:, 1]),
    'XGB_tuned': np.array(y_test_XGB[:, 1]),
    'LR2_with_embeddings': np.array(y_test_LR2),
    'CB7_with_embeddings': np.array(y_test_CB7[:, 1]),
    'CB8_with_embeddings': np.array(y_test_CB8[:, 1]),
    'WeightedEnsemble': np.array(y_test_WE[1]),
}

# Initialize lift chart
fig, ax = plt.subplots(figsize=(8, 8))
ax.set_title("Lift chart for high risk area (top decile)")
ax.set_xlabel('Fraction of high risk readmissions')
ax.set_ylabel('Lift')
plt.xlim(0, 0.15)
plt.ylim(0, 8.0)

# Calculate and plot lift score for each model
for label, predictions in model_predictions.items():
    percentages, gains_adjusted = calculate_gains_and_percentages(predictions, y_true)
    ax.plot(percentages, gains_adjusted, lw=2, label=label)

# Plot random choice baseline
ax.plot([0, 1], [1, 1], 'k--', lw=2, label='random choice')
ax.legend(loc='upper right')
plt.show()

# Merge predicted an true values into a DataFrame
df = pd.DataFrame({
    'PredProbability': y_test_CB1[:,1],
    'TrueValue': y_test
})

# Sort by Probability in descending order
df_sorted = df.sort_values(by='PredProbability', ascending=False)
df_sorted.head()

# Creating a 'Percentile' column based on quantile-based discretization into 100 bins
df_sorted['Percentile'] = pd.qcut(df_sorted['PredProbability'], 100, labels=range(1, 101))

# Calculate means of predicted and true values for the top 10 percentiles
percentile_means = df_sorted.groupby('Percentile', observed=True).mean().tail(10)

# Printing the results for the top 10 percentiles
print(percentile_means[::-1])  # This reverses the order to start with the highest probabilities

            PredProbability  TrueValue
Percentile                            
100                0.632874   0.569444
99                 0.466461   0.527778
98                 0.405462   0.338028
97                 0.368264   0.361111
96                 0.339783   0.375000
95                 0.314150   0.253521
94                 0.290905   0.347222
93                 0.271439   0.267606
92                 0.256960   0.305556
91                 0.245194   0.277778

Hyperparameter	Value
nan_mode	Min
eval_metric	AUC
combinations_ctr	['Borders:CtrBorderCount=15:CtrBorderType=Uniform:TargetBorderCount=1:TargetBorderType=MinEntropy:Prior=0/1:Prior=0.5/1:Prior=1/1', 'Counter:CtrBorderCount=15:CtrBorderType=Uniform:Prior=0/1']
iterations	1000
sampling_frequency	PerTree
fold_permutation_block	0
leaf_estimation_method	Newton
random_score_type	NormalWithModelSizeDecrease
counter_calc_method	SkipTest
grow_policy	SymmetricTree
penalties_coefficient	1
boosting_type	Plain
model_shrink_mode	Constant
feature_border_type	GreedyLogSum
ctr_leaf_count_limit	18446744073709551615
bayesian_matrix_reg	0.100000
one_hot_max_size	2
eval_fraction	0
force_unit_auto_pair_weights	False
l2_leaf_reg	3
random_strength	1
rsm	1
boost_from_average	False
max_ctr_complexity	4
model_size_reg	0.500000
simple_ctr	['Borders:CtrBorderCount=15:CtrBorderType=Uniform:TargetBorderCount=1:TargetBorderType=MinEntropy:Prior=0/1:Prior=0.5/1:Prior=1/1', 'Counter:CtrBorderCount=15:CtrBorderType=Uniform:Prior=0/1']
pool_metainfo_options	{'tags': {}}
subsample	0.800000
use_best_model	False
class_names	[0, 1]
random_seed	42
depth	6
ctr_target_border_count	1
posterior_sampling	False
has_time	False
store_all_simple_ctr	False
border_count	254
classes_count	0
auto_class_weights	None
sparse_features_conflict_fraction	0
leaf_estimation_backtracking	AnyImprovement
best_model_min_trees	1
model_shrink_rate	0
min_data_in_leaf	1
loss_function	Logloss
learning_rate	0.046101
score_function	Cosine
task_type	CPU
leaf_estimation_iterations	10
bootstrap_type	MVS
max_leaves	64
permutation_count	4

	param_learning_rate	param_depth	mean_test_score	rank_test_score
4	0.020941	6	0.692364	1
1	0.032482	6	0.692006	2
5	0.026312	4	0.691834	3
0	0.036038	6	0.691131	4
2	0.026132	4	0.690959	5

	param_learning_rate	param_num_leaves	param_subsample	param_colsample_bytree	mean_test_score	rank_test_score
11	0.007744	50	0.495177	0.241025	0.690353	1
1	0.003802	40	0.058084	0.596850	0.690204	2
18	0.002320	50	0.275999	0.356753	0.688115	3
13	0.003574	32	0.546710	0.755361	0.687634	4
2	0.006054	50	0.650888	0.866176	0.687624	5

	param_learning_rate	param_max_depth	param_subsample	param_colsample_bytree	mean_test_score	rank_test_score
4	0.006945	9	0.611653	0.181825	0.689892	1
8	0.005434	8	0.680308	0.592415	0.688936	2
7	0.015141	9	0.514234	0.090606	0.688423	3
0	0.027464	8	0.779691	0.374540	0.688003	4
2	0.014680	9	0.650888	0.866176	0.687881	5

category	count
Diabetes mellitus without complication	1874
Congestive heart failure; nonhypertensive	1186
Coronary atherosclerosis and other heart disease	1057

	race	gender	age	weight	admission_type_id	discharge_disposition_id	admission_source_id	time_in_hospital	payer_code	medical_specialty	num_lab_procedures	num_procedures	num_medications	number_outpatient	diag_1	diag_2	diag_3	number_diagnoses	max_glu_serum	A1Cresult	metformin	repaglinide	nateglinide	chlorpropamide	glimepiride	acetohexamide	glipizide	glyburide	tolbutamide	pioglitazone	rosiglitazone	acarbose	miglitol	troglitazone	tolazamide	examide	citoglipton	insulin	glyburide-metformin	glipizide-metformin	glimepiride-pioglitazone	metformin-rosiglitazone	metformin-pioglitazone	change	diabetesMed	TARGET	group_diag_1	group_diag_2	group_diag_3
47523	Asian	Female	[50-60)	?	3	6	1	1	?	Orthopedics	18	4	6	0	722	250	401	4	None	None	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	0	Musculoskeletal	Diabetes	Circulatory
8582	Caucasian	Male	[40-50)	?	2	6	4	8	?	?	45	6	53	0	410	414	401	7	None	None	No	No	No	No	No	No	No	Up	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	Ch	Yes	0	Circulatory	Circulatory	Circulatory
42024	Caucasian	Female	[70-80)	?	1	3	7	3	MC	?	47	0	13	0	25012	427	276	9	None	None	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	0	Diabetes	Circulatory	Other
15219	AfricanAmerican	Female	[20-30)	?	1	1	7	2	?	InternalMedicine	35	2	7	0	25013	280	789	5	None	None	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	Steady	No	No	No	No	No	No	Yes	0	Diabetes	Other	Neoplasms
45649	Caucasian	Female	[60-70)	?	1	2	7	1	?	?	23	0	11	0	410	414	250	5	None	>8	No	No	No	No	No	No	No	Steady	No	No	No	No	No	No	No	No	No	Steady	No	No	No	No	No	Ch	Yes	1	Circulatory	Circulatory	Diabetes
23448	Caucasian	Male	[80-90)	?	3	28	1	1	SP	?	55	0	13	0	294	312	414	5	None	None	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	1	Other	Other	Circulatory
39945	Caucasian	Male	[40-50)	?	1	1	7	4	?	?	56	2	16	1	535	584	25041	9	None	None	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	No	1	Other	Genitourinary	Diabetes

Categorical Feature	Number of Unique Values
diag_3	724
diag_2	678
diag_1	661
medical_specialty	45
payer_code	17
discharge_disposition_id	15
admission_source_id	11
group_diag_3	10
group_diag_1	10
age	10
group_diag_2	10
weight	8
admission_type_id	7
race	6
metformin	4
pioglitazone	4
max_glu_serum	4
A1Cresult	4
insulin	4
repaglinide	4
rosiglitazone	4
glyburide	4
glimepiride	4
glipizide	4
acarbose	3
nateglinide	3
tolbutamide	2
chlorpropamide	2
diabetesMed	2
change	2
gender	2
glyburide-metformin	2
miglitol	2
tolazamide	2
citoglipton	1
glipizide-metformin	1
glimepiride-pioglitazone	1
metformin-rosiglitazone	1
metformin-pioglitazone	1
examide	1
acetohexamide	1
troglitazone	1

	category_id	code	group	chapter	chapter_id	L2code	category
0	0	?	n.a.	n.a	0	n.a.	Not available
1	135	1	Digestive	Diseases of the digestive system	9	c9.1	Intestinal infection
2	1	10	Other	Infectious and parasitic diseases	1	c1.1	Tuberculosis

category	count
Congestive heart failure; nonhypertensive	443
Coronary atherosclerosis and other heart disease	429
Acute cerebrovascular disease	324

category	count
Diabetes mellitus without complication	591
Congestive heart failure; nonhypertensive	438
Fluid and electrolyte disorders	404

category	count
Diabetes mellitus without complication	961
Essential hypertension	474
Fluid and electrolyte disorders	317

discharge_disposition_id	count
1	30258
3	6029
6	5189
18	1921
2	1101
22	1073
5	664
25	494
4	375
7	296
23	181
28	72
8	49
15	28
24	21

medical_specialty	count
?	22642
InternalMedicine	7341
Family/GeneralPractice	3312
Emergency/Trauma	2875
Cardiology	2811
Surgery-General	1528
Orthopedics	891
Orthopedics-Reconstructive	806
Radiologist	576
ObstetricsandGynecology	506
Nephrology	460
Psychiatry	450
Pulmonology	415
Urology	409
Surgery-Cardiovascular/Thoracic	402
Surgery-Neuro	340
Gastroenterology	246
Surgery-Vascular	241
PhysicalMedicineandRehabilitation	153
Oncology	144
Pediatrics	139
Neurology	135
Pediatrics-Endocrinology	131
ZZ	104
Otolaryngology	90
Hematology/Oncology	77
Endocrinology	68
Surgery-Cardiovascular	67
Surgery-Thoracic	66
Pediatrics-CriticalCare	52
Gynecology	46
Psychology	39
Surgeon	35
Podiatry	34
Radiology	29
Ophthalmology	25
Surgery-Plastic	23
Hospitalist	23
InfectiousDiseases	20

payer_code	count
?	20378
MC	13115
HM	2821
BC	2601
SP	2259
MD	1522
CP	1391
UN	1385
CM	913
OG	475
PO	357
DM	256
WC	96
CH	92
OT	42
SI	26
MP	22

age	count
[70-80)	11623
[60-70)	10736
[50-60)	8615
[80-90)	7242
[40-50)	4885
[30-40)	2003
[90-100)	1282
[20-30)	844
[10-20)	392
[0-10)	129

admission_type_id	count
1	24036
3	10056
2	8731
6	2641
5	2047
8	222
99	18

	layers	dropout_rate	optimizer	batch_size	epochs	val_loss	val_auc
0	(30, 10)	0.3	adam	32	20	0.359087	0.692090
1	(30, 10)	0.1	adam	64	10	0.358597	0.689873
2	10	0.1	adam	32	10	0.358947	0.688083
3	(20, 5)	0.3	adam	64	20	0.361231	0.687680
4	(30, 20, 5)	0.1	adam	32	10	0.360035	0.685874
5	(30, 20, 10)	0.3	adam	64	10	0.361238	0.685116
6	10	0.5	adam	64	10	0.361394	0.684720
7	10	0.1	adam	32	20	0.360505	0.684214
8	30	0.5	adam	64	20	0.360139	0.684154
9	10	0.5	adam	32	20	0.361033	0.684061
10	(20, 10)	0.1	adam	64	10	0.360023	0.683964
11	(30, 20, 10)	0.1	adam	64	10	0.359962	0.683775

	model	score_test	score_val	eval_metric	pred_time_test	pred_time_val	fit_time	pred_time_test_marginal	pred_time_val_marginal	fit_time_marginal	stack_level	can_infer	fit_order
0	LightGBM_BAG_L2	0.706243	0.698350	roc_auc	36.435412	50.441354	1027.245120	1.341141	0.623350	8.168139	2	True	24
1	LightGBM_r96_BAG_L2	0.705856	0.701887	roc_auc	37.057245	50.668730	1025.605496	1.962975	0.850726	6.528515	2	True	39
2	CatBoost_BAG_L2	0.705774	0.703741	roc_auc	36.082140	51.708746	1072.608280	0.987869	1.890742	53.531299	2	True	27
3	CatBoost_r177_BAG_L2	0.705515	0.703445	roc_auc	36.313934	51.491967	1073.612370	1.219664	1.673963	54.535389	2	True	34
4	RandomForestEntr_BAG_L1	0.705087	0.684488	roc_auc	0.566349	8.355005	12.145532	0.566349	8.355005	12.145532	1	True	6
5	LightGBM_r131_BAG_L2	0.704889	0.697478	roc_auc	36.857112	50.634384	1030.057138	1.762841	0.816380	10.980157	2	True	36
6	WeightedEnsemble_L2	0.704813	0.705625	roc_auc	13.358419	25.202521	601.281765	0.042000	0.000000	1.471262	2	True	22
7	LightGBMXT_BAG_L2	0.704788	0.700562	roc_auc	36.396357	50.646163	1026.403245	1.302086	0.828159	7.326264	2	True	23
8	CatBoost_r9_BAG_L2	0.704519	0.700336	roc_auc	36.573185	51.716511	1174.326463	1.478915	1.898508	155.249482	2	True	38
9	XGBoost_BAG_L2	0.704185	0.695643	roc_auc	37.234827	50.486380	1053.603121	2.140556	0.668376	34.526139	2	True	31
10	LightGBMLarge_BAG_L1	0.704128	0.686622	roc_auc	1.641981	0.673794	29.963329	1.641981	0.673794	29.963329	1	True	13
11	RandomForestGini_BAG_L1	0.704010	0.682610	roc_auc	0.662482	9.003107	10.533237	0.662482	9.003107	10.533237	1	True	5
12	ExtraTreesEntr_BAG_L2	0.703902	0.693596	roc_auc	35.602563	57.126904	1030.688657	0.508293	7.308901	11.611676	2	True	29
13	LightGBMLarge_BAG_L2	0.703586	0.694109	roc_auc	36.754710	50.329182	1048.688829	1.660439	0.511178	29.611848	2	True	33
14	LightGBM_r131_BAG_L1	0.703490	0.694592	roc_auc	2.158012	1.291697	10.845987	2.158012	1.291697	10.845987	1	True	16

	PredProbability	TrueValue
34138	0.888056	1
33582	0.885106	1
45194	0.857854	1
33042	0.853814	1
32855	0.846220	0

Author: Chow Chun Mun¶

Practical Examination in Actuarial Data Science Immersion, 13 April – 13 May 2025¶

Part II (Python)¶

Diabetes Hospital Readmission: Focus on Neural Networks and Embeddings¶

Task PT0: Setting Up a Development Environment [Learning objectives: 5.1 — Points: 1]]¶

Task PT1: Creating a Good Benchmark Model as Simply as Possible [Learning objectives: 5.1 — Points: 4] ¶

Part A: Quick & Easy

1. Developing an Initial Baseline Machine Learning Model

1.1 Setting Up the Modeling Environment

1.2 Load and examine the input data

1.3 Replacing aka. detecting missing categorical values

1.4 Partitioning data into training, validation, and test subsets

1.5 Training a default CatBoost classifier (baseline model)

Task PT2: Logistic Regression and Feature Analysis [Learning objectives: 2.2, 5.1 — Points: 10] ¶

Part B: Generating Insights from Data and Modeling

2. Logistic Regression: A Classical Approach to Classification

2.1 Preparing the data: Imputing missing numbers with median

2.2 Logistic regression implementation: Fitting and performance evaluation

2.3 Data imputation impact: Re-evaluating CatBoost with median replacement

3. Exploratory Data Analysis and Feature Engineering

3.1 Analysis of categorical features

3.2 Analysis of numerical features

6. Data Preprocessing: Encoding, Scaling, and Subsampling

6.1 Encoding categorical data and scaling numerical values

6.2 Do encoding and scaling affect the benchmark model?

6.3 Subsampling the training data

6.4 Does subsampling affect the benchmark model?

6.5 Model evaluation and comparison

Task PT3: Adjusting and Extending Model Optimization [Learning objectives: 5.1 — Points: 3]] ¶

Part C: Tuning and Applying Machine Learning Models

9. Hyperparameter Optimization in CatBoost

10. HP-Tuning LightGBM and XGBoost

10.1 LightGBM: HP-tuning and evaluation

10.2 XGBoost: HP-tuning and evaluation

Task PT4: Creating DXG Diagnosis Groups and Preparing for Embedding Generation ¶

Task PT5: Creating Stable Diagnosis Embeddings with a High-Performance Neural Network¶

Task PT6: Joining Embeddings and Using Them in Modeling [Learning objectives: 5.1 — Points: 15]¶

Task PT7: AutoML, Model Evaluation, and Application [Learning objectives: 5.1 — Points: 16]¶

11. Model Evaluation and Application

11.1 Model evaluation: Validation and test set

11.2 High risk prediction